ContextStream

resmi

Memori persisten dan pencarian semantik untuk asisten pengkodean AI di seluruh sesi

Dokumentasi

Docs · Telusuri bagian

01 · Wizard penyiapan

Satu perintah, terkonfigurasi penuh.

Jalankan satu perintah untuk mengautentikasi (login browser/perangkat), membuat kunci API, menghasilkan aturan, dan menulis konfigurasi MCP yang benar untuk alat Anda (termasuk skema servers VS Code).

terminal · terminal · macOS / Linux

curl -fsSL https://contextstream.io/scripts/mcp.sh | bash

terminal · powershell · Windows

irm https://contextstream.io/scripts/mcp.ps1 | iex

Sudah terinstal? Jalankan ulang wizard penyiapan

terminal · terminal · jalankan ulang

contextstream-mcp setup

Apa yang dilakukannya: menulis konfigurasi MCP (VS Code servers, Cursor/Cline/dll mcpServers), menghasilkan aturan (Standar/Ditingkatkan), dan dapat menautkan proyek ke ruang kerja.

Perangkat alat terkonsolidasi: Wizard mengonfigurasi ~11 alat domain terkonsolidasi secara default (~75% pengurangan token). Opsional: mode router (~2 alat-meta, paling ringkas). Lihat katalog alat lengkap.

Pratinjau perubahan tanpa menulis file: contextstream-mcp setup --dry-run

Alur kerja tim

Memori, keterampilan, dan pemunculan konteks bersama.

Akun tim mendapatkan lebih dari sekadar proyek bersama. Selama contextstream-mcp setup, wizard mendeteksi kemampuan tim dan memunculkan tips ruang kerja. Di editor Anda, setiap panggilan session(action="context") dapat menyertakan rekomendasi tim, isyarat tata kelola, sinyal prioritas, dan artefak tertaut.

Pilih ruang kerja bersama

Tautkan setiap repo ke ruang kerja tim selama penyiapan agar pengindeksan, keputusan, dan tiket tetap selaras.

Bagikan keterampilan tim

skill(action="share", scope="team") menerbitkan alur kerja yang dapat digunakan kembali yang secara otomatis dicocokkan oleh rekan tim di session(action="context").

Ganti cakupan eksekusi

Akun konteks ganda menggunakan --account-mode=team|personal|auto atau session set_account_mode di MCP.

Lacak pekerjaan dengan entitas

Tiket, serah terima, insiden, dan rilis menggunakan referensi terindeks — tahan lama di seluruh sesi dan rekan tim.

Remote terhosting adalah default. Biner MCP lokal hanya untuk pemulihan — atur CONTEXTSTREAM_ALLOW_LOCAL_MCP=1 saat benar-benar diperlukan. Lihat penyiapan tim untuk undangan/peran dan daftar periksa pasca-login.

Pintasan CLI

Perintah non-interaktif (CI & penyegaran).

Jalankan ini tanpa wizard interaktif — ideal setelah peningkatan, orientasi tim, rotasi kredensial, atau perubahan ruang kerja. Juga dimunculkan di contextstream-mcp --help.

Perintah	Kapan digunakan
contextstream-mcp update-hooks --scope=global	Setelah peningkatan atau bergabung dengan ruang kerja tim — segarkan hook PreToolUse/UserPromptSubmit.
contextstream-mcp update-rules --scope=all	Regenerasi .cursorrules / CLAUDE.md / AGENTS.md dengan panduan alur kerja tim terbaru.
contextstream-mcp update-configs --scope=global	Tulis ulang konfigurasi MCP setelah perubahan kunci API atau ruang kerja.
contextstream-mcp migrate-remote --scope=all	Konversi konfigurasi stdio lokal lama ke transport remote terhosting.
contextstream-mcp detect-editors --format=json	Skrip editor mana yang terinstal (bootstrap/CI).
contextstream-mcp generate-configs --transport=remote --preauth	Keluarkan payload konfigurasi JSON tanpa menulis file.
contextstream-mcp configure --transcripts=on --scope=all	Atur default tangkapan transkrip secara non-interaktif.

terminal · mode akun · tim vs pribadi

# Default: follow account (auto)
contextstream-mcp --account-mode=auto

# Force team-scoped reads/writes
contextstream-mcp --account-mode=team

# Or set once in shell profile:
export CONTEXTSTREAM_ACCOUNT_MODE=team

02 · Konfigurasi manual

Konfigurasi per-klien.

Gunakan format yang benar per klien (VS Code menggunakan servers; banyak klien lain menggunakan mcpServers).

Perangkat alat terkonsolidasi: Secara default, server mengekspos ~11 alat domain terkonsolidasi (~75% pengurangan token vs alat granular lama). Untuk alat yang lebih sedikit, tambahkan "CONTEXTSTREAM_PROGRESSIVE_MODE": "true" ke blok env untuk ~2 alat-meta router. Lihat katalog alat lengkap.

Langsung ke alat Anda

Cursor / VS Code Windsurf Codex CLI OpenCode CLI Claude Code Claude Desktop Cline Kilo Code Roo Code Antigravity

Apa itu MCP?

Protokol terbuka untuk AI.

Model Context Protocol (MCP) adalah standar terbuka yang memungkinkan asisten AI terhubung ke alat dan sumber data eksternal. Dengan server MCP ContextStream, alat AI Anda dapat:

Mengingat percakapan dan keputusan di seluruh sesi
Mencari basis kode dan dokumentasi Anda secara semantik
Membangun dan menanyakan grafik pengetahuan
Berbagi konteks antara alat AI yang berbeda

Bahasa alami

Tanya saja. AI yang menangani alatnya.

Anda tidak perlu menghafal nama alat atau memanggilnya secara langsung. Cukup jelaskan apa yang Anda inginkan dalam bahasa Inggris biasa dan asisten AI Anda akan menggunakan alat yang tepat secara otomatis.

Tanya saja secara alami

· "ringkasan sesi"
· "apa yang kita putuskan tentang auth?"
· "ingat kita menggunakan PostgreSQL"
· "cari kode pembayaran"

AI menangani sisanya

· Menemukan konteks yang relevan secara otomatis
· Mengingat keputusan masa lalu saat dibutuhkan
· Menyimpan informasi penting ke memori
· Mencari kode dan dokumentasi

Contoh · "ringkasan sesi"

Natural language example: typing 'session summary' and the AI automatically uses context_smart

AI memahami maksud Anda dan memanggil alat ContextStream yang sesuai di belakang layar.

Prasyarat

Apa yang Anda butuhkan.

Akun ContextStream (wizard penyiapan dapat membuat kunci API melalui login browser).

Perkaya konteks

Integrasi GitHub + Slack

MCP memberi AI Anda memori persisten. Menghubungkan GitHub dan Slack membuat memori itu jauh lebih kaya — AI Anda dapat secara otomatis merujuk PR, isu, dan diskusi tim saat menjawab pertanyaan.

Pengayaan konteks otomatis

Saat Anda memanggil context_smart atau session_smart_search, isu GitHub, PR, dan diskusi Slack yang relevan secara otomatis disertakan. Tidak perlu alat tambahan.

GitHubSync isu, PR, rilis, dan komentar. Keputusan secara otomatis diekstrak dari diskusi.SlackSync saluran dan utas. Percakapan dengan keterlibatan tinggi dinilai dan diprioritaskan.

Contoh prompt

· "Apa yang kita putuskan tentang autentikasi?" — menemukan keputusan dari isu GitHub + utas Slack
· "Tunjukkan aktivitas terbaru di sistem pembayaran" — memunculkan PR, isu, dan diskusi tim
· "Pelajaran apa yang kita dapat dari pemadaman terakhir?" — mengambil wawasan dari Slack dan GitHub
· "Beri saya ringkasan mingguan aktivitas GitHub" — menggunakan integration(provider="github", action="summary", ...)
· "Cari semua integrasi untuk diskusi migrasi basis data" — menggunakan integration(provider="all", action="search", ...)
· "Beri saya ringkasan tim mingguan di semua sumber" — menggunakan integration(provider="all", action="summary", ...)

Referensi cepat tindakan alat integrasi

Gunakan integration(provider="github|slack|all", action="...")

GitHub (provider="github")

action="stats" — Statistik dan status sinkronisasi
action="search" — Cari dengan filter status/jangka waktu
action="activity" — Umpan aktivitas (filter hari)
action="knowledge" — Keputusan/pelajaran yang diekstrak
action="summary" — Ringkasan mingguan/bulanan
action="repos" — Daftar repo yang disinkronkan
action="issues" — Daftar isu/PR

Slack (provider="slack")

action="stats" — Statistik dan status sinkronisasi
action="search" — Cari dengan filter saluran/jangka waktu
action="discussions" — Utas dengan keterlibatan tinggi
action="knowledge" — Keputusan/pelajaran yang diekstrak
action="summary" — Ringkasan mingguan/bulanan
action="channels" — Daftar saluran yang disinkronkan

Lintas sumber (provider="all")

action="status" — Periksa status sinkronisasi dan kesehatan semua integrasi yang terhubung
action="search" — Cari di semua integrasi yang terhubung dalam satu kueri
action="summary" — Ringkasan aktivitas terpadu di semua sumber (filter hari)
action="knowledge" — Dapatkan keputusan, pelajaran, dan wawasan dari semua sumber

Klien · Cursor / VS Code

Cursor / VS Code

Cursor dan VS Code menggunakan skema konfigurasi MCP yang berbeda. Cursor masih menggunakan proses MCP lokal, tetapi VS Code/Copilot sekarang dapat menggunakan MCP ContextStream terhosting langsung melalui HTTP.

terminal · .cursor/mcp.json (proyek) atau ~/.cursor/mcp.json (global)

{
  "mcpServers": {
    "contextstream": {
      "command": "contextstream-mcp",
      "env": {
        "CONTEXTSTREAM_API_URL": "https://api.contextstream.io",
        "CONTEXTSTREAM_API_KEY": "your_api_key"
      }
    }
  }
}

Direkomendasikan untuk VS Code / Copilot: instal remote satu klik

Instal MCP ContextStream terhosting dan biarkan VS Code menangani OAuth pada penggunaan pertama. Tidak perlu biner lokal atau kunci API di .vscode/mcp.json.

Instal di VS Code Dokumen MCP VS Code

terminal · .vscode/mcp.json (MCP asli VS Code, remote)

{
  "servers": {
    "contextstream": {
      "type": "http",
      "url": "https://mcp.contextstream.io/mcp?default_context_mode=fast"
    }
  }
}

Lebih suka baris perintah? Tambahkan server remote dengan code --add-mcp

terminal · terminal

code --add-mcp '{"name":"contextstream","type":"http","url":"https://mcp.contextstream.io/mcp?default_context_mode=fast"}'

Pada penggunaan pertama, VS Code akan meminta untuk mengotorisasi ContextStream dan kemudian menyelesaikan penyiapan secara otomatis.

Self-hosted? Arahkan konfigurasi remote yang sama ke URL gateway MCP Anda sendiri alih-alih https://mcp.contextstream.io/mcp?default_context_mode=fast.

Klien · OpenCode CLI

OpenCode CLI

Untuk menggunakan ContextStream dengan OpenCode CLI, tambahkan konfigurasi server MCP ke file ~/.config/opencode/opencode.json Anda (atau opencode.json di root proyek Anda):

terminal · ~/.config/opencode/opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "contextstream": {
      "type": "local",
      "command": ["contextstream-mcp"],
      "enabled": true,
      "environment": {
        "CONTEXTSTREAM_API_URL": "https://api.contextstream.io",
        "CONTEXTSTREAM_API_KEY": "your_api_key"
      }
    }
  }
}

Setelah mengedit konfigurasi, mulai ulang OpenCode agar dapat memuat server MCP ContextStream.

Klien · Codex CLI

Codex CLI

Untuk menggunakan ContextStream dengan Codex CLI, tambahkan konfigurasi server MCP ke file ~/.codex/config.toml Anda:

terminal · ~/.codex/config.toml

[mcp_servers.contextstream]
command = "contextstream-mcp"
args = []

[mcp_servers.contextstream.env]
CONTEXTSTREAM_API_URL = "https://api.contextstream.io"
CONTEXTSTREAM_API_KEY = "your_api_key"

Setelah mengedit konfigurasi, mulai ulang Codex agar dapat memuat server MCP ContextStream.

Klien · Claude Code

Claude Code (CLI)

Tambahkan ContextStream ke Claude Code dengan menjalankan perintah ini dari direktori proyek Anda:

terminal · terminal

claude mcp add --transport stdio contextstream --env CONTEXTSTREAM_API_URL=https://api.contextstream.io --env CONTEXTSTREAM_API_KEY=your_api_key -- contextstream-mcp

Peringatan Windows (Windows asli, bukan WSL): gunakan cmd /c contextstream-mcp setelah --.

Alternatif: add-json (stdio)

terminal · terminal · add-json

claude mcp add-json contextstream \
'{"type":"stdio","command":"contextstream-mcp","args":[],"env":{"CONTEXTSTREAM_API_URL":"https://api.contextstream.io","CONTEXTSTREAM_API_KEY":"your_api_key"}}'

Tips: untuk penyiapan tim, lebih baik gunakan file .mcp.json yang dikomit (cakupan proyek) daripada menyematkan kunci di riwayat shell.

Ini menambahkan ContextStream ke ~/.claude.json di bawah jalur proyek Anda, membuatnya tersedia saat bekerja di direktori itu.

Opsi cakupan MCP

· Lokal (default): Pribadi untuk Anda, hanya proyek saat ini → disimpan di ~/.claude.json di bawah jalur proyek.
· Pengguna (--scope user): Pribadi untuk Anda, semua proyek → disimpan di ~/.claude.json secara global.
· Proyek (--scope project): Dibagikan dengan tim → disimpan di .mcp.json di root proyek (komit ke git).

Untuk berbagi tim, gunakan cakupan proyek untuk membuat file .mcp.json yang dapat dikomit ke git:

terminal · .mcp.json (cakupan proyek)

{
  "mcpServers": {
    "contextstream": {
      "command": "contextstream-mcp",
      "env": {
        "CONTEXTSTREAM_API_URL": "https://api.contextstream.io",
        "CONTEXTSTREAM_API_KEY": "your_api_key"
      }
    }
  }
}

Setelah menambahkan server MCP, mulai ulang Claude Code agar perubahan diterapkan. Verifikasi server dimuat dengan claude mcp list. Untuk server bercakupan proyek dari .mcp.json, Claude Code akan meminta persetujuan pada penggunaan pertama.

Klien · Claude Desktop

Claude Desktop (aplikasi GUI)

Tambahkan ContextStream ke aplikasi Claude Desktop:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Windows: %APPDATA%\Claude\claude_desktop_config.json

terminal · claude_desktop_config.json

{
  "mcpServers": {
    "contextstream": {
      "command": "contextstream-mcp",
      "env": {
        "CONTEXTSTREAM_API_URL": "https://api.contextstream.io",
        "CONTEXTSTREAM_API_KEY": "your_api_key"
      }
    }
  }
}

Setelah mengedit konfigurasi, keluar dan mulai ulang Claude Desktop agar perubahan diterapkan.

Klien · Windsurf

Windsurf

Tambahkan ContextStream ke Windsurf dengan mengedit file konfigurasi MCP global:

Konfigurasi: ~/.codeium/windsurf/mcp_config.json

terminal · ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "contextstream": {
      "command": "contextstream-mcp",
      "env": {
        "CONTEXTSTREAM_API_URL": "https://api.contextstream.io",
        "CONTEXTSTREAM_API_KEY": "your_api_key"
      }
    }
  }
}

File aturan

· Global: ~/.codeium/windsurf/memories/global_rules.md
· Proyek: .windsurf/rules/contextstream.md

Windsurf mendukung hook untuk penegakan otomatis aturan ContextStream. Setelah mengedit konfigurasi, mulai ulang Windsurf agar perubahan diterapkan.

Klien · Cline

Cline

Tambahkan ContextStream ke konfigurasi MCP Cline Anda. Klik ikon MCP Servers di Cline, pilih tab "Configure", lalu klik "Configure MCP Servers" untuk mengedit:

terminal · cline_mcp_settings.json

{
  "mcpServers": {
    "contextstream": {
      "command": "contextstream-mcp",
      "env": {
        "CONTEXTSTREAM_API_URL": "https://api.contextstream.io",
        "CONTEXTSTREAM_API_KEY": "your_api_key"
      }
    }
  }
}

Setelah mengedit konfigurasi, mulai ulang Cline agar perubahan diterapkan. Anda juga dapat menggunakan alwaysAllow untuk menyetujui otomatis alat tertentu.

Klien · Kilo Code

Kilo Code

Tambahkan ContextStream ke konfigurasi MCP Kilo Code Anda. Anda dapat mengonfigurasi server MCP secara global atau per-proyek:

Global: Klik Settings → MCP Servers → Installed → Edit Global MCP untuk membuka mcp_settings.json

Proyek: .kilocode/mcp.json di root proyek Anda

terminal · .kilocode/mcp.json (atau mcp_settings.json)

{
  "mcpServers": {
    "contextstream": {
      "command": "contextstream-mcp",
      "env": {
        "CONTEXTSTREAM_API_URL": "https://api.contextstream.io",
        "CONTEXTSTREAM_API_KEY": "your_api_key"
      }
    }
  }
}

Konfigurasi tingkat proyek lebih diutamakan daripada konfigurasi global. Mulai ulang Kilo Code setelah mengedit.

Klien · Roo Code

Roo Code

Tambahkan ContextStream ke konfigurasi MCP Roo Code Anda. Anda dapat mengonfigurasi server MCP secara global atau per-proyek:

Global: Klik ikon pengaturan → Edit Global MCP untuk membuka mcp_settings.json

Proyek: .roo/mcp.json di root proyek Anda

terminal · .roo/mcp.json (atau mcp_settings.json)

{
  "mcpServers": {
    "contextstream": {
      "command": "contextstream-mcp",
      "env": {
        "CONTEXTSTREAM_API_URL": "https://api.contextstream.io",
        "CONTEXTSTREAM_API_KEY": "your_api_key"
      }
    }
  }
}

Konfigurasi tingkat proyek lebih diutamakan daripada konfigurasi global. Mulai ulang Roo Code setelah mengedit.

Klien · Antigravity

Antigravity (Google)

Antigravity menggunakan file .mcp.json bercakupan proyek dengan format yang sama seperti Cursor/Claude Desktop:

terminal · .mcp.json (root proyek)

{
  "mcpServers": {
    "contextstream": {
      "command": "contextstream-mcp",
      "env": {
        "CONTEXTSTREAM_API_URL": "https://api.contextstream.io",
        "CONTEXTSTREAM_API_KEY": "your_api_key"
      }
    }
  }
}

Pengguna Windows: gunakan wrapper cmd

terminal · .mcp.json (Windows)

{
  "mcpServers": {
    "contextstream": {
      "command": "cmd",
      "args": ["/c", "contextstream-mcp"],
      "env": {
        "CONTEXTSTREAM_API_URL": "https://api.contextstream.io",
        "CONTEXTSTREAM_API_KEY": "your_api_key"
      }
    }
  }
}

File aturan

· Global: ~/.gemini/GEMINI.md
· Ruang Kerja: .agent/rules/contextstream.md

Akses melalui menu "..." → "MCP Servers" → "View raw config" untuk memverifikasi konfigurasi Anda. Mulai ulang Antigravity setelah mengedit.

Aturan editor

Tingkatkan penggunaan ContextStream otomatis

Direkomendasikan

Fitur konteks otomatis ContextStream memuat konteks ruang kerja secara otomatis saat Anda menggunakan alat ContextStream apa pun. Namun, asisten AI mungkin tidak selalu secara proaktif menyimpan keputusan atau mengingat konteks masa lalu. Menambahkan aturan AI editor meningkatkan konsistensi dan memastikan AI secara otomatis menangkap keputusan, preferensi, dan konteks penting di seluruh percakapan Anda.

Penting: Konvensi penamaan alat MCP

Alat AI yang berbeda menggunakan konvensi penamaan yang berbeda untuk alat MCP. Menggunakan format yang salah akan menyebabkan alat tidak ditemukan.

Alat AI	Format	Contoh
Claude Code	mcp____	mcp__contextstream__session_init
Codex CLI / OpenCode CLI	(nama mentah)	session_init
Cursor / Windsurf / Cline	(nama mentah)	session_init
Kilo Code / Roo Code	(nama mentah)	session_init

Ringkasan: Hanya Claude Code yang menggunakan awalan mcp__contextstream__. Semua alat lain menggunakan nama alat mentah.

Anda dapat menambahkan aturan ContextStream di dua tingkat: Global (berlaku untuk semua proyek) atau Proyek (berlaku untuk satu proyek).

Aturan global (semua proyek)

Tambahkan ini sekali dan mereka akan berlaku untuk setiap proyek secara otomatis:

Editor	Lokasi aturan global
Cursor	Settings → General → Rules for AI
Windsurf	~/.codeium/windsurf/memories/global_rules.md
Cline	~/Documents/Cline/Rules/
Kilo Code	~/.kilocode/rules/
Roo Code	~/.roo/rules/
Claude Code	~/.claude/CLAUDE.md
Codex CLI	~/.codex/AGENTS.md (global) atau folder induk (mis. ~/dev/AGENTS.md)
OpenCode CLI	~/.config/opencode/AGENTS.md

Aturan proyek (proyek tunggal)

Tambahkan ini ke proyek tertentu. Untuk aturan berbasis folder Cursor, atur mode aktivasi ke "Always On" agar aturan selalu aktif.

Editor	Lokasi aturan proyek
Cursor	.cursorrules atau .cursor/rules/*.mdc
Windsurf	.windsurf/rules/contextstream.md
Cline	file .clinerules atau folder .clinerules/
Kilo Code	.kilocode/rules/
Roo Code	.roo/rules/contextstream.md atau folder .roo/rules/
Claude Code	CLAUDE.md di root proyek
Codex CLI	AGENTS.md di root proyek
OpenCode CLI	AGENTS.md di root proyek
Aider	.aider.conf.yml di root proyek

Mode aktivasi (Cursor, Kilo Code & Roo Code)

Saat menggunakan aturan berbasis folder (mis., .cursor/rules/), setiap file aturan memiliki mode aktivasi:

Always On — Selalu aktif (direkomendasikan untuk ContextStream)
Manual — Hanya saat Anda @mention aturan
Model Decision — AI memutuskan berdasarkan deskripsi
Glob — Aktif untuk pola file yang cocok

Aturan global dan file tingkat root (.cursorrules) selalu aktif.

Lebih suka wizard penyiapan? Jalankan dulu. Jika tidak, tambahkan aturan standar ini secara manual. Contoh untuk setiap editor:

Claude Code

Buat file CLAUDE.md di root proyek Anda atau tambahkan ke ~/.claude/CLAUDE.md global Anda:

terminal · CLAUDE.md (Standar)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `mcp__contextstream__search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `mcp__contextstream__context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x (Hooks Enforced)

Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

### Required Every Message

| Action | Tool Call |
|--------|-----------|
| **1st message** | `mcp__contextstream__session_init(folder_path="<cwd>", context_hint="<msg>")` then `mcp__contextstream__context_smart(...)` |
| **2nd+ messages** | `mcp__contextstream__context_smart(user_message="<msg>", format="minified", max_tokens=400)` |
| **Code search** | `mcp__contextstream__search(mode="hybrid", query="...")` — BEFORE any local tools |
| **Save decisions** | `mcp__contextstream__session(action="capture", event_type="decision", ...)` |

### Search Modes

| Mode | Use Case |
|------|----------|
| `hybrid` | General code mcp__contextstream__search (default) |
| `keyword` | Exact symbol/string match |
| `exhaustive` | Find ALL matches (grep-like) |
| `semantic` | Conceptual questions |

### Why ContextStream First?

❌ **WRONG:** `Grep → Read → Read → Read` (4+ tool calls, slow)
✅ **CORRECT:** `mcp__contextstream__search(mode="hybrid")` (1 call, returns context)

ContextStream search is **indexed** and returns semantic matches + context in ONE call.

### Quick Reference

| Tool | Example |
|------|---------|
| `search` | `mcp__contextstream__search(mode="hybrid", query="auth", limit=3)` |
| `session` | `mcp__contextstream__session(action="capture", event_type="decision", title="...", content="...")` |
| `memory` | `mcp__contextstream__memory(action="list_events", limit=10)` |
| `graph` | `mcp__contextstream__graph(action="dependencies", file_path="...")` |

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `mcp__contextstream__session(action="get_lessons", query="<topic>")`
- On mistakes: `mcp__contextstream__session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

### Plans & Tasks

When user asks for a plan, use ContextStream (not EnterPlanMode):
1. `mcp__contextstream__session(action="capture_plan", title="...", steps=[...])`
2. `mcp__contextstream__memory(action="create_task", title="...", plan_id="<id>")`

Full docs: https://contextstream.io/docs/mcp/tools

Tampilkan aturan yang ditingkatkan (verbose)

terminal · CLAUDE.md (Ditingkatkan)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `mcp__contextstream__search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `mcp__contextstream__context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x Integration (Enhanced)

You have access to ContextStream MCP tools for persistent memory and context.
v0.4.x uses **~11 consolidated domain tools** for ~75% token reduction vs previous versions.
Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

## TL;DR - REQUIRED EVERY MESSAGE

| Message | What to Call |
|---------|--------------|
| **1st message** | `mcp__contextstream__session_init(folder_path="...", context_hint="<user's message>")`, then `mcp__contextstream__context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **2nd+ messages** | `mcp__contextstream__context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **🔍 ANY code search** | `mcp__contextstream__search(mode="hybrid", query="...")` — ALWAYS before Glob/Grep/Search/Read |
| **Before risky/non-trivial work** | `mcp__contextstream__session(action="get_lessons", query="<topic>")` |
| **After completing task** | `mcp__contextstream__session(action="capture", event_type="decision", ...)` - MUST capture |
| **User frustration/correction** | `mcp__contextstream__session(action="capture_lesson", ...)` - MUST capture lessons |

**NO EXCEPTIONS.** Do not skip even if you think you have enough context.

**First message rule:** After `session_init`, always call `context_smart` before any other tool or response.

**Context Pack (Pro+):** If enabled, use `mcp__contextstream__context_smart(..., mode="pack", distill=true)` for code/file queries. If unavailable or disabled, omit `mode` and proceed with standard `context_smart` (the API will fall back).

**Tool naming:** Use the exact tool names exposed by your MCP client. Claude Code typically uses `mcp__<server>__<tool>` where `<server>` matches your MCP config (often `contextstream`). If a tool call fails with "No such tool available", refresh rules and match the tool list.

---

## Consolidated Domain Tools Architecture

v0.4.x consolidates ~58 individual tools into ~11 domain tools with action/mode dispatch:

### Standalone Tools (Always Call)
- **`session_init`** - Initialize session with workspace detection + context
- **`context_smart`** - Semantic search for relevant context (CALL EVERY MESSAGE, including immediately after `session_init`)

### Domain Tools (Use action/mode parameter)

| Domain | Actions/Modes | Example |
|--------|---------------|---------|
| **`search`** | mode: semantic, hybrid, keyword, pattern, exhaustive, refactor | `mcp__contextstream__search(mode="hybrid", query="auth implementation", limit=3)` |
| **`session`** | action: capture, capture_lesson, get_lessons, recall, remember, user_context, summary, compress, delta, smart_search, decision_trace | `mcp__contextstream__session(action="capture", event_type="decision", title="Use JWT", content="...")` |
| **`memory`** | action: create_event, get_event, update_event, delete_event, list_events, distill_event, create_node, get_node, update_node, delete_node, list_nodes, supersede_node, search, decisions, timeline, summary | `mcp__contextstream__memory(action="list_events", limit=10)` |
| **`graph`** | action: dependencies, impact, call_path, related, path, decisions, ingest, circular_dependencies, unused_code, contradictions | `mcp__contextstream__graph(action="impact", symbol_name="AuthService")` |
| **`project`** | action: list, get, create, update, index, overview, statistics, files, index_status, ingest_local | `mcp__contextstream__project(action="statistics")` |
| **`workspace`** | action: list, get, associate, bootstrap | `mcp__contextstream__workspace(action="list")` |
| **`reminder`** | action: list, active, create, snooze, complete, dismiss | `mcp__contextstream__reminder(action="active")` |
| **`integration`** | provider: slack/github/all; action: status, search, stats, activity, contributors, knowledge, summary, channels, discussions, sync_users, repos, issues | `mcp__contextstream__integration(provider="github", action="search", query="...")` |
| **`help`** | action: tools, auth, version, editor_rules, enable_bundle | `mcp__contextstream__help(action="tools")` |

---

### Why context_smart is Required (Even After session_init)

**Common mistake:** "session_init already gave me context, I don't need context_smart"

**This is WRONG. Here's why:**
- `session_init` returns the last ~10 items **BY TIME** (chronological)
- `context_smart` **SEARCHES** for items **RELEVANT to THIS message** (semantic)

**Example failure:**
- User asks: "how should I implement authentication?"
- Auth decisions were made 20 conversations ago
- `session_init` won't have it (too old, not in recent 10)
- `context_smart` FINDS it via semantic search

**Without context_smart, you WILL miss relevant older context.**

---

### Search & Code Intelligence (ContextStream-first)

⚠️ **STOP: Before using Search/Glob/Grep/Read/Explore** → Call `mcp__contextstream__search(mode="hybrid")` FIRST. Use local tools ONLY if ContextStream returns 0 results.

**❌ WRONG workflow (wastes tokens, slow):**

Grep "function" → Baca file1.ts → Baca file2.ts → Baca file3.ts → akhirnya mengerti


**✅ CORRECT workflow (fast, complete):**

mcp__contextstream__search(mode="hybrid", query="implementasi fungsi") → selesai (hasil termasuk konteks)


**Why?** ContextStream search returns semantic matches + context + file locations in ONE call. Local tools require multiple round-trips.

**Search Mode Selection:**

| Need | Mode | Example |
|------|------|---------|
| Find code by meaning | `hybrid` | "authentication logic", "error handling" |
| Exact string/symbol | `keyword` | "UserAuthService", "API_KEY" |
| File patterns | `pattern` | "*.sql", "test_*.py" |
| ALL matches (grep-like) | `exhaustive` | "TODO", "FIXME" (find all occurrences) |
| Symbol renaming | `refactor` | "oldFunctionName" (word-boundary matching) |
| Conceptual search | `semantic` | "how does caching work" |

**Token Efficiency:** Use `output_format` to reduce response size:
- `full` (default): Full content for understanding code
- `paths`: File paths only (80% token savings) - use for file listings
- `minimal`: Compact format (60% savings) - use for refactoring
- `count`: Match counts only (90% savings) - use for quick checks

---

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `mcp__contextstream__session(action="get_lessons", query="<topic>")`
- On mistakes: `mcp__contextstream__session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

---

### Plans & Tasks

When user asks to create a plan or implementation roadmap:
1. Create plan: `mcp__contextstream__session(action="capture_plan", title="Plan Title", description="...", goals=["goal1", "goal2"], steps=[{id: "1", title: "Step 1", order: 1}, ...])`
2. Get plan_id from response, then create tasks: `mcp__contextstream__memory(action="create_task", title="Task Title", plan_id="<plan_id>", priority="high|medium|low", description="...")`

To manage existing plans/tasks:
- List plans: `mcp__contextstream__session(action="list_plans")`
- Get plan with tasks: `mcp__contextstream__session(action="get_plan", plan_id="<uuid>", include_tasks=true)`
- List tasks: `mcp__contextstream__memory(action="list_tasks", plan_id="<uuid>")` or `mcp__contextstream__memory(action="list_tasks")` for all
- Update task status: `mcp__contextstream__memory(action="update_task", task_id="<uuid>", task_status="pending|in_progress|completed|blocked")`

---

### Rules Update Notices

- If you see **[RULES_NOTICE]**, update rules via `mcp__contextstream__generate_rules(overwrite_existing=true)` (preserves custom rules).
- If you see **[VERSION_NOTICE]**, tell the user to update MCP using the provided command.

See full documentation: https://contextstream.io/docs/mcp/tools

Codex CLI / OpenCode CLI

Buat file AGENTS.md di root proyek Anda (aturan proyek) atau di ~/.codex/AGENTS.md (Codex global) / ~/.config/opencode/AGENTS.md (OpenCode global):

Nama alat Codex / OpenCode vs Claude

Codex / OpenCode menggunakan nama alat MCP mentah (mis., session_init). Claude Code menggunakan nama alat dengan namespace (mis., mcp__contextstream__session_init). Gunakan format yang benar untuk alat AI Anda.

terminal · AGENTS.md (Standar)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x (Hooks Enforced)

Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

### Required Every Message

| Action | Tool Call |
|--------|-----------|
| **1st message** | `session_init(folder_path="<cwd>", context_hint="<msg>")` then `context_smart(...)` |
| **2nd+ messages** | `context_smart(user_message="<msg>", format="minified", max_tokens=400)` |
| **Code search** | `search(mode="hybrid", query="...")` — BEFORE any local tools |
| **Save decisions** | `session(action="capture", event_type="decision", ...)` |

### Search Modes

| Mode | Use Case |
|------|----------|
| `hybrid` | General code search (default) |
| `keyword` | Exact symbol/string match |
| `exhaustive` | Find ALL matches (grep-like) |
| `semantic` | Conceptual questions |

### Why ContextStream First?

❌ **WRONG:** `Grep → Read → Read → Read` (4+ tool calls, slow)
✅ **CORRECT:** `search(mode="hybrid")` (1 call, returns context)

ContextStream search is **indexed** and returns semantic matches + context in ONE call.

### Quick Reference

| Tool | Example |
|------|---------|
| `search` | `search(mode="hybrid", query="auth", limit=3)` |
| `session` | `session(action="capture", event_type="decision", title="...", content="...")` |
| `memory` | `memory(action="list_events", limit=10)` |
| `graph` | `graph(action="dependencies", file_path="...")` |

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `session(action="get_lessons", query="<topic>")`
- On mistakes: `session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

### Plans & Tasks

When user asks for a plan, use ContextStream (not EnterPlanMode):
1. `session(action="capture_plan", title="...", steps=[...])`
2. `memory(action="create_task", title="...", plan_id="<id>")`

Full docs: https://contextstream.io/docs/mcp/tools

Tampilkan aturan yang ditingkatkan (verbose)

terminal · AGENTS.md (Ditingkatkan)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x Integration (Enhanced)

You have access to ContextStream MCP tools for persistent memory and context.
v0.4.x uses **~11 consolidated domain tools** for ~75% token reduction vs previous versions.
Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

## TL;DR - REQUIRED EVERY MESSAGE

| Message | What to Call |
|---------|--------------|
| **1st message** | `session_init(folder_path="...", context_hint="<user's message>")`, then `context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **2nd+ messages** | `context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **🔍 ANY code search** | `search(mode="hybrid", query="...")` — ALWAYS before Glob/Grep/Search/Read |
| **Before risky/non-trivial work** | `session(action="get_lessons", query="<topic>")` |
| **After completing task** | `session(action="capture", event_type="decision", ...)` - MUST capture |
| **User frustration/correction** | `session(action="capture_lesson", ...)` - MUST capture lessons |

**NO EXCEPTIONS.** Do not skip even if you think you have enough context.

**First message rule:** After `session_init`, always call `context_smart` before any other tool or response.

**Context Pack (Pro+):** If enabled, use `context_smart(..., mode="pack", distill=true)` for code/file queries. If unavailable or disabled, omit `mode` and proceed with standard `context_smart` (the API will fall back).

**Tool naming:** Use the exact tool names exposed by your MCP client. Claude Code typically uses `mcp__<server>__<tool>` where `<server>` matches your MCP config (often `contextstream`). If a tool call fails with "No such tool available", refresh rules and match the tool list.

---

## Consolidated Domain Tools Architecture

v0.4.x consolidates ~58 individual tools into ~11 domain tools with action/mode dispatch:

### Standalone Tools (Always Call)
- **`session_init`** - Initialize session with workspace detection + context
- **`context_smart`** - Semantic search for relevant context (CALL EVERY MESSAGE, including immediately after `session_init`)

### Domain Tools (Use action/mode parameter)

| Domain | Actions/Modes | Example |
|--------|---------------|---------|
| **`search`** | mode: semantic, hybrid, keyword, pattern, exhaustive, refactor | `search(mode="hybrid", query="auth implementation", limit=3)` |
| **`session`** | action: capture, capture_lesson, get_lessons, recall, remember, user_context, summary, compress, delta, smart_search, decision_trace | `session(action="capture", event_type="decision", title="Use JWT", content="...")` |
| **`memory`** | action: create_event, get_event, update_event, delete_event, list_events, distill_event, create_node, get_node, update_node, delete_node, list_nodes, supersede_node, search, decisions, timeline, summary | `memory(action="list_events", limit=10)` |
| **`graph`** | action: dependencies, impact, call_path, related, path, decisions, ingest, circular_dependencies, unused_code, contradictions | `graph(action="impact", symbol_name="AuthService")` |
| **`project`** | action: list, get, create, update, index, overview, statistics, files, index_status, ingest_local | `project(action="statistics")` |
| **`workspace`** | action: list, get, associate, bootstrap | `workspace(action="list")` |
| **`reminder`** | action: list, active, create, snooze, complete, dismiss | `reminder(action="active")` |
| **`integration`** | provider: slack/github/all; action: status, search, stats, activity, contributors, knowledge, summary, channels, discussions, sync_users, repos, issues | `integration(provider="github", action="search", query="...")` |
| **`help`** | action: tools, auth, version, editor_rules, enable_bundle | `help(action="tools")` |

---

### Why context_smart is Required (Even After session_init)

**Common mistake:** "session_init already gave me context, I don't need context_smart"

**This is WRONG. Here's why:**
- `session_init` returns the last ~10 items **BY TIME** (chronological)
- `context_smart` **SEARCHES** for items **RELEVANT to THIS message** (semantic)

**Example failure:**
- User asks: "how should I implement authentication?"
- Auth decisions were made 20 conversations ago
- `session_init` won't have it (too old, not in recent 10)
- `context_smart` FINDS it via semantic search

**Without context_smart, you WILL miss relevant older context.**

---

### Search & Code Intelligence (ContextStream-first)

⚠️ **STOP: Before using Search/Glob/Grep/Read/Explore** → Call `search(mode="hybrid")` FIRST. Use local tools ONLY if ContextStream returns 0 results.

**❌ WRONG workflow (wastes tokens, slow):**

Grep "function" → Baca file1.ts → Baca file2.ts → Baca file3.ts → akhirnya mengerti


**✅ CORRECT workflow (fast, complete):**

search(mode="hybrid", query="implementasi fungsi") → selesai (hasil termasuk konteks)


**Why?** ContextStream search returns semantic matches + context + file locations in ONE call. Local tools require multiple round-trips.

**Search Mode Selection:**

| Need | Mode | Example |
|------|------|---------|
| Find code by meaning | `hybrid` | "authentication logic", "error handling" |
| Exact string/symbol | `keyword` | "UserAuthService", "API_KEY" |
| File patterns | `pattern` | "*.sql", "test_*.py" |
| ALL matches (grep-like) | `exhaustive` | "TODO", "FIXME" (find all occurrences) |
| Symbol renaming | `refactor` | "oldFunctionName" (word-boundary matching) |
| Conceptual search | `semantic` | "how does caching work" |

**Token Efficiency:** Use `output_format` to reduce response size:
- `full` (default): Full content for understanding code
- `paths`: File paths only (80% token savings) - use for file listings
- `minimal`: Compact format (60% savings) - use for refactoring
- `count`: Match counts only (90% savings) - use for quick checks

---

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `session(action="get_lessons", query="<topic>")`
- On mistakes: `session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

---

### Plans & Tasks

When user asks to create a plan or implementation roadmap:
1. Create plan: `session(action="capture_plan", title="Plan Title", description="...", goals=["goal1", "goal2"], steps=[{id: "1", title: "Step 1", order: 1}, ...])`
2. Get plan_id from response, then create tasks: `memory(action="create_task", title="Task Title", plan_id="<plan_id>", priority="high|medium|low", description="...")`

To manage existing plans/tasks:
- List plans: `session(action="list_plans")`
- Get plan with tasks: `session(action="get_plan", plan_id="<uuid>", include_tasks=true)`
- List tasks: `memory(action="list_tasks", plan_id="<uuid>")` or `memory(action="list_tasks")` for all
- Update task status: `memory(action="update_task", task_id="<uuid>", task_status="pending|in_progress|completed|blocked")`

---

### Rules Update Notices

- If you see **[RULES_NOTICE]**, update rules via `generate_rules(overwrite_existing=true)` (preserves custom rules).
- If you see **[VERSION_NOTICE]**, tell the user to update MCP using the provided command.

See full documentation: https://contextstream.io/docs/mcp/tools

Windsurf

Buat file .windsurf/rules/contextstream.md di root proyek Anda atau tambahkan ke ~/.codeium/windsurf/memories/global_rules.md global Anda:

terminal · .windsurf/rules/contextstream.md (Standar)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x (Hooks Enforced)

Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

### Required Every Message

| Action | Tool Call |
|--------|-----------|
| **1st message** | `session_init(folder_path="<cwd>", context_hint="<msg>")` then `context_smart(...)` |
| **2nd+ messages** | `context_smart(user_message="<msg>", format="minified", max_tokens=400)` |
| **Code search** | `search(mode="hybrid", query="...")` — BEFORE any local tools |
| **Save decisions** | `session(action="capture", event_type="decision", ...)` |

### Search Modes

| Mode | Use Case |
|------|----------|
| `hybrid` | General code search (default) |
| `keyword` | Exact symbol/string match |
| `exhaustive` | Find ALL matches (grep-like) |
| `semantic` | Conceptual questions |

### Why ContextStream First?

❌ **WRONG:** `Grep → Read → Read → Read` (4+ tool calls, slow)
✅ **CORRECT:** `search(mode="hybrid")` (1 call, returns context)

ContextStream search is **indexed** and returns semantic matches + context in ONE call.

### Quick Reference

| Tool | Example |
|------|---------|
| `search` | `search(mode="hybrid", query="auth", limit=3)` |
| `session` | `session(action="capture", event_type="decision", title="...", content="...")` |
| `memory` | `memory(action="list_events", limit=10)` |
| `graph` | `graph(action="dependencies", file_path="...")` |

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `session(action="get_lessons", query="<topic>")`
- On mistakes: `session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

### Plans & Tasks

When user asks for a plan, use ContextStream (not EnterPlanMode):
1. `session(action="capture_plan", title="...", steps=[...])`
2. `memory(action="create_task", title="...", plan_id="<id>")`

Full docs: https://contextstream.io/docs/mcp/tools

Tampilkan aturan yang ditingkatkan (verbose)

terminal · .windsurf/rules/contextstream.md (Ditingkatkan)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x Integration (Enhanced)

You have access to ContextStream MCP tools for persistent memory and context.
v0.4.x uses **~11 consolidated domain tools** for ~75% token reduction vs previous versions.
Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

## TL;DR - REQUIRED EVERY MESSAGE

| Message | What to Call |
|---------|--------------|
| **1st message** | `session_init(folder_path="...", context_hint="<user's message>")`, then `context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **2nd+ messages** | `context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **🔍 ANY code search** | `search(mode="hybrid", query="...")` — ALWAYS before Glob/Grep/Search/Read |
| **Before risky/non-trivial work** | `session(action="get_lessons", query="<topic>")` |
| **After completing task** | `session(action="capture", event_type="decision", ...)` - MUST capture |
| **User frustration/correction** | `session(action="capture_lesson", ...)` - MUST capture lessons |

**NO EXCEPTIONS.** Do not skip even if you think you have enough context.

**First message rule:** After `session_init`, always call `context_smart` before any other tool or response.

**Context Pack (Pro+):** If enabled, use `context_smart(..., mode="pack", distill=true)` for code/file queries. If unavailable or disabled, omit `mode` and proceed with standard `context_smart` (the API will fall back).

**Tool naming:** Use the exact tool names exposed by your MCP client. Claude Code typically uses `mcp__<server>__<tool>` where `<server>` matches your MCP config (often `contextstream`). If a tool call fails with "No such tool available", refresh rules and match the tool list.

---

## Consolidated Domain Tools Architecture

v0.4.x consolidates ~58 individual tools into ~11 domain tools with action/mode dispatch:

### Standalone Tools (Always Call)
- **`session_init`** - Initialize session with workspace detection + context
- **`context_smart`** - Semantic search for relevant context (CALL EVERY MESSAGE, including immediately after `session_init`)

### Domain Tools (Use action/mode parameter)

| Domain | Actions/Modes | Example |
|--------|---------------|---------|
| **`search`** | mode: semantic, hybrid, keyword, pattern, exhaustive, refactor | `search(mode="hybrid", query="auth implementation", limit=3)` |
| **`session`** | action: capture, capture_lesson, get_lessons, recall, remember, user_context, summary, compress, delta, smart_search, decision_trace | `session(action="capture", event_type="decision", title="Use JWT", content="...")` |
| **`memory`** | action: create_event, get_event, update_event, delete_event, list_events, distill_event, create_node, get_node, update_node, delete_node, list_nodes, supersede_node, search, decisions, timeline, summary | `memory(action="list_events", limit=10)` |
| **`graph`** | action: dependencies, impact, call_path, related, path, decisions, ingest, circular_dependencies, unused_code, contradictions | `graph(action="impact", symbol_name="AuthService")` |
| **`project`** | action: list, get, create, update, index, overview, statistics, files, index_status, ingest_local | `project(action="statistics")` |
| **`workspace`** | action: list, get, associate, bootstrap | `workspace(action="list")` |
| **`reminder`** | action: list, active, create, snooze, complete, dismiss | `reminder(action="active")` |
| **`integration`** | provider: slack/github/all; action: status, search, stats, activity, contributors, knowledge, summary, channels, discussions, sync_users, repos, issues | `integration(provider="github", action="search", query="...")` |
| **`help`** | action: tools, auth, version, editor_rules, enable_bundle | `help(action="tools")` |

---

### Why context_smart is Required (Even After session_init)

**Common mistake:** "session_init already gave me context, I don't need context_smart"

**This is WRONG. Here's why:**
- `session_init` returns the last ~10 items **BY TIME** (chronological)
- `context_smart` **SEARCHES** for items **RELEVANT to THIS message** (semantic)

**Example failure:**
- User asks: "how should I implement authentication?"
- Auth decisions were made 20 conversations ago
- `session_init` won't have it (too old, not in recent 10)
- `context_smart` FINDS it via semantic search

**Without context_smart, you WILL miss relevant older context.**

---

### Search & Code Intelligence (ContextStream-first)

⚠️ **STOP: Before using Search/Glob/Grep/Read/Explore** → Call `search(mode="hybrid")` FIRST. Use local tools ONLY if ContextStream returns 0 results.

**❌ WRONG workflow (wastes tokens, slow):**

Grep "function" → Baca file1.ts → Baca file2.ts → Baca file3.ts → akhirnya mengerti


**✅ CORRECT workflow (fast, complete):**

search(mode="hybrid", query="implementasi fungsi") → selesai (hasil termasuk konteks)


**Why?** ContextStream search returns semantic matches + context + file locations in ONE call. Local tools require multiple round-trips.

**Search Mode Selection:**

| Need | Mode | Example |
|------|------|---------|
| Find code by meaning | `hybrid` | "authentication logic", "error handling" |
| Exact string/symbol | `keyword` | "UserAuthService", "API_KEY" |
| File patterns | `pattern` | "*.sql", "test_*.py" |
| ALL matches (grep-like) | `exhaustive` | "TODO", "FIXME" (find all occurrences) |
| Symbol renaming | `refactor` | "oldFunctionName" (word-boundary matching) |
| Conceptual search | `semantic` | "how does caching work" |

**Token Efficiency:** Use `output_format` to reduce response size:
- `full` (default): Full content for understanding code
- `paths`: File paths only (80% token savings) - use for file listings
- `minimal`: Compact format (60% savings) - use for refactoring
- `count`: Match counts only (90% savings) - use for quick checks

---

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `session(action="get_lessons", query="<topic>")`
- On mistakes: `session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

---

### Plans & Tasks

When user asks to create a plan or implementation roadmap:
1. Create plan: `session(action="capture_plan", title="Plan Title", description="...", goals=["goal1", "goal2"], steps=[{id: "1", title: "Step 1", order: 1}, ...])`
2. Get plan_id from response, then create tasks: `memory(action="create_task", title="Task Title", plan_id="<plan_id>", priority="high|medium|low", description="...")`

To manage existing plans/tasks:
- List plans: `session(action="list_plans")`
- Get plan with tasks: `session(action="get_plan", plan_id="<uuid>", include_tasks=true)`
- List tasks: `memory(action="list_tasks", plan_id="<uuid>")` or `memory(action="list_tasks")` for all
- Update task status: `memory(action="update_task", task_id="<uuid>", task_status="pending|in_progress|completed|blocked")`

---

### Rules Update Notices

- If you see **[RULES_NOTICE]**, update rules via `generate_rules(overwrite_existing=true)` (preserves custom rules).
- If you see **[VERSION_NOTICE]**, tell the user to update MCP using the provided command.

See full documentation: https://contextstream.io/docs/mcp/tools

Kilo Code

Buat file Markdown di .kilocode/rules/:

terminal · .kilocode/rules/contextstream.md (Standar)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x (Hooks Enforced)

Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

### Required Every Message

| Action | Tool Call |
|--------|-----------|
| **1st message** | `session_init(folder_path="<cwd>", context_hint="<msg>")` then `context_smart(...)` |
| **2nd+ messages** | `context_smart(user_message="<msg>", format="minified", max_tokens=400)` |
| **Code search** | `search(mode="hybrid", query="...")` — BEFORE any local tools |
| **Save decisions** | `session(action="capture", event_type="decision", ...)` |

### Search Modes

| Mode | Use Case |
|------|----------|
| `hybrid` | General code search (default) |
| `keyword` | Exact symbol/string match |
| `exhaustive` | Find ALL matches (grep-like) |
| `semantic` | Conceptual questions |

### Why ContextStream First?

❌ **WRONG:** `Grep → Read → Read → Read` (4+ tool calls, slow)
✅ **CORRECT:** `search(mode="hybrid")` (1 call, returns context)

ContextStream search is **indexed** and returns semantic matches + context in ONE call.

### Quick Reference

| Tool | Example |
|------|---------|
| `search` | `search(mode="hybrid", query="auth", limit=3)` |
| `session` | `session(action="capture", event_type="decision", title="...", content="...")` |
| `memory` | `memory(action="list_events", limit=10)` |
| `graph` | `graph(action="dependencies", file_path="...")` |

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `session(action="get_lessons", query="<topic>")`
- On mistakes: `session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

### Plans & Tasks

When user asks for a plan, use ContextStream (not EnterPlanMode):
1. `session(action="capture_plan", title="...", steps=[...])`
2. `memory(action="create_task", title="...", plan_id="<id>")`

Full docs: https://contextstream.io/docs/mcp/tools

Tampilkan aturan yang ditingkatkan (verbose)

terminal · .kilocode/rules/contextstream.md (Ditingkatkan)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x Integration (Enhanced)

You have access to ContextStream MCP tools for persistent memory and context.
v0.4.x uses **~11 consolidated domain tools** for ~75% token reduction vs previous versions.
Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

## TL;DR - REQUIRED EVERY MESSAGE

| Message | What to Call |
|---------|--------------|
| **1st message** | `session_init(folder_path="...", context_hint="<user's message>")`, then `context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **2nd+ messages** | `context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **🔍 ANY code search** | `search(mode="hybrid", query="...")` — ALWAYS before Glob/Grep/Search/Read |
| **Before risky/non-trivial work** | `session(action="get_lessons", query="<topic>")` |
| **After completing task** | `session(action="capture", event_type="decision", ...)` - MUST capture |
| **User frustration/correction** | `session(action="capture_lesson", ...)` - MUST capture lessons |

**NO EXCEPTIONS.** Do not skip even if you think you have enough context.

**First message rule:** After `session_init`, always call `context_smart` before any other tool or response.

**Context Pack (Pro+):** If enabled, use `context_smart(..., mode="pack", distill=true)` for code/file queries. If unavailable or disabled, omit `mode` and proceed with standard `context_smart` (the API will fall back).

**Tool naming:** Use the exact tool names exposed by your MCP client. Claude Code typically uses `mcp__<server>__<tool>` where `<server>` matches your MCP config (often `contextstream`). If a tool call fails with "No such tool available", refresh rules and match the tool list.

---

## Consolidated Domain Tools Architecture

v0.4.x consolidates ~58 individual tools into ~11 domain tools with action/mode dispatch:

### Standalone Tools (Always Call)
- **`session_init`** - Initialize session with workspace detection + context
- **`context_smart`** - Semantic search for relevant context (CALL EVERY MESSAGE, including immediately after `session_init`)

### Domain Tools (Use action/mode parameter)

| Domain | Actions/Modes | Example |
|--------|---------------|---------|
| **`search`** | mode: semantic, hybrid, keyword, pattern, exhaustive, refactor | `search(mode="hybrid", query="auth implementation", limit=3)` |
| **`session`** | action: capture, capture_lesson, get_lessons, recall, remember, user_context, summary, compress, delta, smart_search, decision_trace | `session(action="capture", event_type="decision", title="Use JWT", content="...")` |
| **`memory`** | action: create_event, get_event, update_event, delete_event, list_events, distill_event, create_node, get_node, update_node, delete_node, list_nodes, supersede_node, search, decisions, timeline, summary | `memory(action="list_events", limit=10)` |
| **`graph`** | action: dependencies, impact, call_path, related, path, decisions, ingest, circular_dependencies, unused_code, contradictions | `graph(action="impact", symbol_name="AuthService")` |
| **`project`** | action: list, get, create, update, index, overview, statistics, files, index_status, ingest_local | `project(action="statistics")` |
| **`workspace`** | action: list, get, associate, bootstrap | `workspace(action="list")` |
| **`reminder`** | action: list, active, create, snooze, complete, dismiss | `reminder(action="active")` |
| **`integration`** | provider: slack/github/all; action: status, search, stats, activity, contributors, knowledge, summary, channels, discussions, sync_users, repos, issues | `integration(provider="github", action="search", query="...")` |
| **`help`** | action: tools, auth, version, editor_rules, enable_bundle | `help(action="tools")` |

---

### Why context_smart is Required (Even After session_init)

**Common mistake:** "session_init already gave me context, I don't need context_smart"

**This is WRONG. Here's why:**
- `session_init` returns the last ~10 items **BY TIME** (chronological)
- `context_smart` **SEARCHES** for items **RELEVANT to THIS message** (semantic)

**Example failure:**
- User asks: "how should I implement authentication?"
- Auth decisions were made 20 conversations ago
- `session_init` won't have it (too old, not in recent 10)
- `context_smart` FINDS it via semantic search

**Without context_smart, you WILL miss relevant older context.**

---

### Search & Code Intelligence (ContextStream-first)

⚠️ **STOP: Before using Search/Glob/Grep/Read/Explore** → Call `search(mode="hybrid")` FIRST. Use local tools ONLY if ContextStream returns 0 results.

**❌ WRONG workflow (wastes tokens, slow):**

Grep "function" → Baca file1.ts → Baca file2.ts → Baca file3.ts → akhirnya mengerti


**✅ CORRECT workflow (fast, complete):**

search(mode="hybrid", query="implementasi fungsi") → selesai (hasil termasuk konteks)


**Why?** ContextStream search returns semantic matches + context + file locations in ONE call. Local tools require multiple round-trips.

**Search Mode Selection:**

| Need | Mode | Example |
|------|------|---------|
| Find code by meaning | `hybrid` | "authentication logic", "error handling" |
| Exact string/symbol | `keyword` | "UserAuthService", "API_KEY" |
| File patterns | `pattern` | "*.sql", "test_*.py" |
| ALL matches (grep-like) | `exhaustive` | "TODO", "FIXME" (find all occurrences) |
| Symbol renaming | `refactor` | "oldFunctionName" (word-boundary matching) |
| Conceptual search | `semantic` | "how does caching work" |

**Token Efficiency:** Use `output_format` to reduce response size:
- `full` (default): Full content for understanding code
- `paths`: File paths only (80% token savings) - use for file listings
- `minimal`: Compact format (60% savings) - use for refactoring
- `count`: Match counts only (90% savings) - use for quick checks

---

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `session(action="get_lessons", query="<topic>")`
- On mistakes: `session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

---

### Plans & Tasks

When user asks to create a plan or implementation roadmap:
1. Create plan: `session(action="capture_plan", title="Plan Title", description="...", goals=["goal1", "goal2"], steps=[{id: "1", title: "Step 1", order: 1}, ...])`
2. Get plan_id from response, then create tasks: `memory(action="create_task", title="Task Title", plan_id="<plan_id>", priority="high|medium|low", description="...")`

To manage existing plans/tasks:
- List plans: `session(action="list_plans")`
- Get plan with tasks: `session(action="get_plan", plan_id="<uuid>", include_tasks=true)`
- List tasks: `memory(action="list_tasks", plan_id="<uuid>")` or `memory(action="list_tasks")` for all
- Update task status: `memory(action="update_task", task_id="<uuid>", task_status="pending|in_progress|completed|blocked")`

---

### Rules Update Notices

- If you see **[RULES_NOTICE]**, update rules via `generate_rules(overwrite_existing=true)` (preserves custom rules).
- If you see **[VERSION_NOTICE]**, tell the user to update MCP using the provided command.

See full documentation: https://contextstream.io/docs/mcp/tools

Cline

Buat file .clinerules di root proyek Anda atau gunakan folder .clinerules/:

terminal · .clinerules (Standar)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x (Hooks Enforced)

Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

### Required Every Message

| Action | Tool Call |
|--------|-----------|
| **1st message** | `session_init(folder_path="<cwd>", context_hint="<msg>")` then `context_smart(...)` |
| **2nd+ messages** | `context_smart(user_message="<msg>", format="minified", max_tokens=400)` |
| **Code search** | `search(mode="hybrid", query="...")` — BEFORE any local tools |
| **Save decisions** | `session(action="capture", event_type="decision", ...)` |

### Search Modes

| Mode | Use Case |
|------|----------|
| `hybrid` | General code search (default) |
| `keyword` | Exact symbol/string match |
| `exhaustive` | Find ALL matches (grep-like) |
| `semantic` | Conceptual questions |

### Why ContextStream First?

❌ **WRONG:** `Grep → Read → Read → Read` (4+ tool calls, slow)
✅ **CORRECT:** `search(mode="hybrid")` (1 call, returns context)

ContextStream search is **indexed** and returns semantic matches + context in ONE call.

### Quick Reference

| Tool | Example |
|------|---------|
| `search` | `search(mode="hybrid", query="auth", limit=3)` |
| `session` | `session(action="capture", event_type="decision", title="...", content="...")` |
| `memory` | `memory(action="list_events", limit=10)` |
| `graph` | `graph(action="dependencies", file_path="...")` |

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `session(action="get_lessons", query="<topic>")`
- On mistakes: `session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

### Plans & Tasks

When user asks for a plan, use ContextStream (not EnterPlanMode):
1. `session(action="capture_plan", title="...", steps=[...])`
2. `memory(action="create_task", title="...", plan_id="<id>")`

Full docs: https://contextstream.io/docs/mcp/tools

Tampilkan aturan yang ditingkatkan (verbose)

terminal · .clinerules (Ditingkatkan)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x Integration (Enhanced)

You have access to ContextStream MCP tools for persistent memory and context.
v0.4.x uses **~11 consolidated domain tools** for ~75% token reduction vs previous versions.
Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

## TL;DR - REQUIRED EVERY MESSAGE

| Message | What to Call |
|---------|--------------|
| **1st message** | `session_init(folder_path="...", context_hint="<user's message>")`, then `context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **2nd+ messages** | `context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **🔍 ANY code search** | `search(mode="hybrid", query="...")` — ALWAYS before Glob/Grep/Search/Read |
| **Before risky/non-trivial work** | `session(action="get_lessons", query="<topic>")` |
| **After completing task** | `session(action="capture", event_type="decision", ...)` - MUST capture |
| **User frustration/correction** | `session(action="capture_lesson", ...)` - MUST capture lessons |

**NO EXCEPTIONS.** Do not skip even if you think you have enough context.

**First message rule:** After `session_init`, always call `context_smart` before any other tool or response.

**Context Pack (Pro+):** If enabled, use `context_smart(..., mode="pack", distill=true)` for code/file queries. If unavailable or disabled, omit `mode` and proceed with standard `context_smart` (the API will fall back).

**Tool naming:** Use the exact tool names exposed by your MCP client. Claude Code typically uses `mcp__<server>__<tool>` where `<server>` matches your MCP config (often `contextstream`). If a tool call fails with "No such tool available", refresh rules and match the tool list.

---

## Consolidated Domain Tools Architecture

v0.4.x consolidates ~58 individual tools into ~11 domain tools with action/mode dispatch:

### Standalone Tools (Always Call)
- **`session_init`** - Initialize session with workspace detection + context
- **`context_smart`** - Semantic search for relevant context (CALL EVERY MESSAGE, including immediately after `session_init`)

### Domain Tools (Use action/mode parameter)

| Domain | Actions/Modes | Example |
|--------|---------------|---------|
| **`search`** | mode: semantic, hybrid, keyword, pattern, exhaustive, refactor | `search(mode="hybrid", query="auth implementation", limit=3)` |
| **`session`** | action: capture, capture_lesson, get_lessons, recall, remember, user_context, summary, compress, delta, smart_search, decision_trace | `session(action="capture", event_type="decision", title="Use JWT", content="...")` |
| **`memory`** | action: create_event, get_event, update_event, delete_event, list_events, distill_event, create_node, get_node, update_node, delete_node, list_nodes, supersede_node, search, decisions, timeline, summary | `memory(action="list_events", limit=10)` |
| **`graph`** | action: dependencies, impact, call_path, related, path, decisions, ingest, circular_dependencies, unused_code, contradictions | `graph(action="impact", symbol_name="AuthService")` |
| **`project`** | action: list, get, create, update, index, overview, statistics, files, index_status, ingest_local | `project(action="statistics")` |
| **`workspace`** | action: list, get, associate, bootstrap | `workspace(action="list")` |
| **`reminder`** | action: list, active, create, snooze, complete, dismiss | `reminder(action="active")` |
| **`integration`** | provider: slack/github/all; action: status, search, stats, activity, contributors, knowledge, summary, channels, discussions, sync_users, repos, issues | `integration(provider="github", action="search", query="...")` |
| **`help`** | action: tools, auth, version, editor_rules, enable_bundle | `help(action="tools")` |

---

### Why context_smart is Required (Even After session_init)

**Common mistake:** "session_init already gave me context, I don't need context_smart"

**This is WRONG. Here's why:**
- `session_init` returns the last ~10 items **BY TIME** (chronological)
- `context_smart` **SEARCHES** for items **RELEVANT to THIS message** (semantic)

**Example failure:**
- User asks: "how should I implement authentication?"
- Auth decisions were made 20 conversations ago
- `session_init` won't have it (too old, not in recent 10)
- `context_smart` FINDS it via semantic search

**Without context_smart, you WILL miss relevant older context.**

---

### Search & Code Intelligence (ContextStream-first)

⚠️ **STOP: Before using Search/Glob/Grep/Read/Explore** → Call `search(mode="hybrid")` FIRST. Use local tools ONLY if ContextStream returns 0 results.

**❌ WRONG workflow (wastes tokens, slow):**

Grep "function" → Baca file1.ts → Baca file2.ts → Baca file3.ts → akhirnya mengerti


**✅ CORRECT workflow (fast, complete):**

search(mode="hybrid", query="implementasi fungsi") → selesai (hasil termasuk konteks)


**Why?** ContextStream search returns semantic matches + context + file locations in ONE call. Local tools require multiple round-trips.

**Search Mode Selection:**

| Need | Mode | Example |
|------|------|---------|
| Find code by meaning | `hybrid` | "authentication logic", "error handling" |
| Exact string/symbol | `keyword` | "UserAuthService", "API_KEY" |
| File patterns | `pattern` | "*.sql", "test_*.py" |
| ALL matches (grep-like) | `exhaustive` | "TODO", "FIXME" (find all occurrences) |
| Symbol renaming | `refactor` | "oldFunctionName" (word-boundary matching) |
| Conceptual search | `semantic` | "how does caching work" |

**Token Efficiency:** Use `output_format` to reduce response size:
- `full` (default): Full content for understanding code
- `paths`: File paths only (80% token savings) - use for file listings
- `minimal`: Compact format (60% savings) - use for refactoring
- `count`: Match counts only (90% savings) - use for quick checks

---

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `session(action="get_lessons", query="<topic>")`
- On mistakes: `session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

---

### Plans & Tasks

When user asks to create a plan or implementation roadmap:
1. Create plan: `session(action="capture_plan", title="Plan Title", description="...", goals=["goal1", "goal2"], steps=[{id: "1", title: "Step 1", order: 1}, ...])`
2. Get plan_id from response, then create tasks: `memory(action="create_task", title="Task Title", plan_id="<plan_id>", priority="high|medium|low", description="...")`

To manage existing plans/tasks:
- List plans: `session(action="list_plans")`
- Get plan with tasks: `session(action="get_plan", plan_id="<uuid>", include_tasks=true)`
- List tasks: `memory(action="list_tasks", plan_id="<uuid>")` or `memory(action="list_tasks")` for all
- Update task status: `memory(action="update_task", task_id="<uuid>", task_status="pending|in_progress|completed|blocked")`

---

### Rules Update Notices

- If you see **[RULES_NOTICE]**, update rules via `generate_rules(overwrite_existing=true)` (preserves custom rules).
- If you see **[VERSION_NOTICE]**, tell the user to update MCP using the provided command.

See full documentation: https://contextstream.io/docs/mcp/tools

Roo Code

Buat file .roo/rules/contextstream.md atau gunakan folder .roo/rules/:

terminal · .roo/rules/contextstream.md (Standar)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x (Hooks Enforced)

Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

### Required Every Message

| Action | Tool Call |
|--------|-----------|
| **1st message** | `session_init(folder_path="<cwd>", context_hint="<msg>")` then `context_smart(...)` |
| **2nd+ messages** | `context_smart(user_message="<msg>", format="minified", max_tokens=400)` |
| **Code search** | `search(mode="hybrid", query="...")` — BEFORE any local tools |
| **Save decisions** | `session(action="capture", event_type="decision", ...)` |

### Search Modes

| Mode | Use Case |
|------|----------|
| `hybrid` | General code search (default) |
| `keyword` | Exact symbol/string match |
| `exhaustive` | Find ALL matches (grep-like) |
| `semantic` | Conceptual questions |

### Why ContextStream First?

❌ **WRONG:** `Grep → Read → Read → Read` (4+ tool calls, slow)
✅ **CORRECT:** `search(mode="hybrid")` (1 call, returns context)

ContextStream search is **indexed** and returns semantic matches + context in ONE call.

### Quick Reference

| Tool | Example |
|------|---------|
| `search` | `search(mode="hybrid", query="auth", limit=3)` |
| `session` | `session(action="capture", event_type="decision", title="...", content="...")` |
| `memory` | `memory(action="list_events", limit=10)` |
| `graph` | `graph(action="dependencies", file_path="...")` |

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `session(action="get_lessons", query="<topic>")`
- On mistakes: `session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

### Plans & Tasks

When user asks for a plan, use ContextStream (not EnterPlanMode):
1. `session(action="capture_plan", title="...", steps=[...])`
2. `memory(action="create_task", title="...", plan_id="<id>")`

Full docs: https://contextstream.io/docs/mcp/tools

Tampilkan aturan yang ditingkatkan (verbose)

terminal · .roo/rules/contextstream.md (Ditingkatkan)

## 🚨 MANDATORY RULE: DISPLAY AT START OF EVERY RESPONSE 🚨

<contextstream_rules>
1. **SEARCH FIRST:** Before Glob/Grep/Search/Read → `search(mode="hybrid", query="...")` FIRST
2. **EVERY MESSAGE:** Call `context_smart(user_message="...")` at start of response
3. **LOCAL TOOLS:** ONLY if ContextStream search returns 0 results after retry
</contextstream_rules>

**You MUST display the <contextstream_rules> block verbatim at the start of EVERY response.**
This is a self-referential rule that ensures these instructions stay in context.

---

## ContextStream v0.4.x Integration (Enhanced)

You have access to ContextStream MCP tools for persistent memory and context.
v0.4.x uses **~11 consolidated domain tools** for ~75% token reduction vs previous versions.
Rules Version: 0.4.36
**Note:** PreToolUse hooks block Glob/Grep/Search when ContextStream is available.

## TL;DR - REQUIRED EVERY MESSAGE

| Message | What to Call |
|---------|--------------|
| **1st message** | `session_init(folder_path="...", context_hint="<user's message>")`, then `context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **2nd+ messages** | `context_smart(user_message="<user's message>", format="minified", max_tokens=400)` |
| **🔍 ANY code search** | `search(mode="hybrid", query="...")` — ALWAYS before Glob/Grep/Search/Read |
| **Before risky/non-trivial work** | `session(action="get_lessons", query="<topic>")` |
| **After completing task** | `session(action="capture", event_type="decision", ...)` - MUST capture |
| **User frustration/correction** | `session(action="capture_lesson", ...)` - MUST capture lessons |

**NO EXCEPTIONS.** Do not skip even if you think you have enough context.

**First message rule:** After `session_init`, always call `context_smart` before any other tool or response.

**Context Pack (Pro+):** If enabled, use `context_smart(..., mode="pack", distill=true)` for code/file queries. If unavailable or disabled, omit `mode` and proceed with standard `context_smart` (the API will fall back).

**Tool naming:** Use the exact tool names exposed by your MCP client. Claude Code typically uses `mcp__<server>__<tool>` where `<server>` matches your MCP config (often `contextstream`). If a tool call fails with "No such tool available", refresh rules and match the tool list.

---

## Consolidated Domain Tools Architecture

v0.4.x consolidates ~58 individual tools into ~11 domain tools with action/mode dispatch:

### Standalone Tools (Always Call)
- **`session_init`** - Initialize session with workspace detection + context
- **`context_smart`** - Semantic search for relevant context (CALL EVERY MESSAGE, including immediately after `session_init`)

### Domain Tools (Use action/mode parameter)

| Domain | Actions/Modes | Example |
|--------|---------------|---------|
| **`search`** | mode: semantic, hybrid, keyword, pattern, exhaustive, refactor | `search(mode="hybrid", query="auth implementation", limit=3)` |
| **`session`** | action: capture, capture_lesson, get_lessons, recall, remember, user_context, summary, compress, delta, smart_search, decision_trace | `session(action="capture", event_type="decision", title="Use JWT", content="...")` |
| **`memory`** | action: create_event, get_event, update_event, delete_event, list_events, distill_event, create_node, get_node, update_node, delete_node, list_nodes, supersede_node, search, decisions, timeline, summary | `memory(action="list_events", limit=10)` |
| **`graph`** | action: dependencies, impact, call_path, related, path, decisions, ingest, circular_dependencies, unused_code, contradictions | `graph(action="impact", symbol_name="AuthService")` |
| **`project`** | action: list, get, create, update, index, overview, statistics, files, index_status, ingest_local | `project(action="statistics")` |
| **`workspace`** | action: list, get, associate, bootstrap | `workspace(action="list")` |
| **`reminder`** | action: list, active, create, snooze, complete, dismiss | `reminder(action="active")` |
| **`integration`** | provider: slack/github/all; action: status, search, stats, activity, contributors, knowledge, summary, channels, discussions, sync_users, repos, issues | `integration(provider="github", action="search", query="...")` |
| **`help`** | action: tools, auth, version, editor_rules, enable_bundle | `help(action="tools")` |

---

### Why context_smart is Required (Even After session_init)

**Common mistake:** "session_init already gave me context, I don't need context_smart"

**This is WRONG. Here's why:**
- `session_init` returns the last ~10 items **BY TIME** (chronological)
- `context_smart` **SEARCHES** for items **RELEVANT to THIS message** (semantic)

**Example failure:**
- User asks: "how should I implement authentication?"
- Auth decisions were made 20 conversations ago
- `session_init` won't have it (too old, not in recent 10)
- `context_smart` FINDS it via semantic search

**Without context_smart, you WILL miss relevant older context.**

---

### Search & Code Intelligence (ContextStream-first)

⚠️ **STOP: Before using Search/Glob/Grep/Read/Explore** → Call `search(mode="hybrid")` FIRST. Use local tools ONLY if ContextStream returns 0 results.

**❌ WRONG workflow (wastes tokens, slow):**

Grep "function" → Baca file1.ts → Baca file2.ts → Baca file3.ts → akhirnya mengerti


**✅ CORRECT workflow (fast, complete):**

search(mode="hybrid", query="implementasi fungsi") → selesai (hasil termasuk konteks)


**Why?** ContextStream search returns semantic matches + context + file locations in ONE call. Local tools require multiple round-trips.

**Search Mode Selection:**

| Need | Mode | Example |
|------|------|---------|
| Find code by meaning | `hybrid` | "authentication logic", "error handling" |
| Exact string/symbol | `keyword` | "UserAuthService", "API_KEY" |
| File patterns | `pattern` | "*.sql", "test_*.py" |
| ALL matches (grep-like) | `exhaustive` | "TODO", "FIXME" (find all occurrences) |
| Symbol renaming | `refactor` | "oldFunctionName" (word-boundary matching) |
| Conceptual search | `semantic` | "how does caching work" |

**Token Efficiency:** Use `output_format` to reduce response size:
- `full` (default): Full content for understanding code
- `paths`: File paths only (80% token savings) - use for file listings
- `minimal`: Compact format (60% savings) - use for refactoring
- `count`: Match counts only (90% savings) - use for quick checks

---

### Lessons (Past Mistakes)

- After `session_init`: Check for `lessons` field and apply before work
- Before risky work: `session(action="get_lessons", query="<topic>")`
- On mistakes: `session(action="capture_lesson", title="...", trigger="...", impact="...", prevention="...")`

---

### Plans & Tasks

When user asks to create a plan or implementation roadmap:
1. Create plan: `session(action="capture_plan", title="Plan Title", description="...", goals=["goal1", "goal2"], steps=[{id: "1", title: "Step 1", order: 1}, ...])`
2. Get plan_id from response, then create tasks: `memory(action="create_task", title="Task Title", plan_id="<plan_id>", priority="high|medium|low", description="...")`

To manage existing plans/tasks:
- List plans: `session(action="list_plans")`
- Get plan with tasks: `session(action="get_plan", plan_id="<uuid>", include_tasks=true)`
- List tasks: `memory(action="list_tasks", plan_id="<uuid>")` or `memory(action="list_tasks")` for all
- Update task status: `memory(action="update_task", task_id="<uuid>", task_status="pending|in_progress|completed|blocked")`

---

### Rules Update Notices

- If you see **[RULES_NOTICE]**, update rules via `generate_rules(overwrite_existing=true)` (preserves custom rules).
- If you see **[VERSION_NOTICE]**, tell the user to update MCP using the provided command.

See full documentation: https://contextstream.io/docs/mcp/tools

Hasilkan aturan otomatis

Anda juga dapat meminta AI untuk menghasilkan aturan ini secara otomatis dengan mengatakan: "Gunakan generate_rules untuk membuat aturan ContextStream untuk proyek ini"

Sistem pelajaran

Sistem pelajaran yang dipetik

Belajar dari kesalahan — jangan pernah mengulanginya

Sistem Pelajaran menangkap kesalahan, koreksi, dan frustrasi pengguna sehingga asisten AI tidak pernah mengulangi kesalahan yang sama. Pelajaran secara otomatis dimunculkan dalam respons session_init dan context_smart saat relevan.

Kapan menangkap pelajaran

Pelajaran harus ditangkap secara otomatis ketika salah satu situasi ini terjadi:

Pemicu	Contoh	Keparahan
Masalah produksi	"Situs down karena perubahan itu"	kritis
Frustrasi pengguna	"TIDAK! Sudah kubilang JANGAN lakukan itu", "WTF", caps lock	tinggi
Koreksi	"Itu salah, seharusnya...", "Perbaiki ini"	sedang
Perubahan yang merusak	"Ini merusak tes", "Build gagal"	sedang/tinggi
Preferensi dinyatakan	"Saya lebih suka cara ini", "Selalu lakukan X saja"	rendah

Bidang pelajaran dijelaskan

Bidang	Deskripsi	Contoh
title	Apa yang harus diingat (imperatif)	"Selalu verifikasi aset di git sebelum push"
severity	kritis, tinggi, sedang, rendah	"kritis" untuk masalah produksi
category	alur_kerja, kualitas_kode, verifikasi, komunikasi, spesifik_proyek	"alur_kerja"
trigger	Tindakan apa yang menyebabkan masalah	"Push kode yang merujuk gambar tanpa mengomitnya"
impact	Apa yang salah	"Error 404 produksi - halaman arahan rusak"
prevention	Cara mencegah di masa depan	"Jalankan git status untuk memeriksa file yang tidak terlacak sebelum push"
keywords	Kata kunci untuk pencocokan di konteks mendatang	["git", "gambar", "aset", "push"]

Contoh lengkap

terminal · session_capture_lesson

// User says: "OH COME ON! You pushed the code but the images are missing
// and now the production site shows broken images!"

session_capture_lesson({
  title: "Always verify assets in git before pushing code references",
  severity: "critical",
  category: "workflow",
  trigger: "Pushed code referencing /screenshots/*.png without committing images",
  impact: "Production 404 errors - broken landing page with missing images",
  prevention: "Run 'git status' to check untracked files before pushing code that references static assets",
  keywords: ["git", "images", "assets", "push", "404", "static", "screenshots"]
})

Bagaimana pelajaran dimunculkan

Pelajaran yang ditangkap secara otomatis dikembalikan di sesi mendatang saat relevan:

`session_init`

Pelajaran dengan keparahan tinggi dan kritis dari ruang kerja ini disertakan dalam inisialisasi sesi, memperingatkan AI sebelum dapat membuat kesalahan yang sama.

`context_smart`

Saat AI meminta konteks, pelajaran yang cocok dengan kata kunci kueri disertakan. Mis., menanyakan tentang "git push" memunculkan pelajaran dengan kata kunci "git" atau "push".

Mengambil pelajaran

Gunakan session_get_lessons untuk mengambil dan memfilter pelajaran:

terminal · contoh session_get_lessons

// Get all critical lessons
session_get_lessons({ severity: "critical" })

// Get workflow lessons
session_get_lessons({ category: "workflow" })

// Search for relevant lessons
session_get_lessons({ query: "git push images" })

// Combine filters
session_get_lessons({
  category: "verification",
  severity: "high",
  limit: 5
})

Tips pro: Tambahkan aturan ke editor Anda (lihat bagian Aturan AI Editor di atas) untuk secara otomatis menangkap pelajaran saat pengguna mengungkapkan frustrasi atau koreksi. Ini membangun basis pengetahuan yang mencegah kesalahan berulang.

Katalog alat

Alat MCP.

Lihat referensi alat MCP lengkap (lencana PRO dan contoh penggunaan umum).

Lihat referensi Alat MCPPerangkat alat terkonsolidasi menggunakan ~11 alat domain untuk ~75% pengurangan token vs alat granular lama.Baca

Contoh penggunaan

Apa yang harus ditanyakan.

Setelah terhubung, Anda dapat menanyakan asisten AI hal-hal seperti:

"Ingat bahwa kita memutuskan untuk menggunakan PostgreSQL untuk basis data"

"Apa keputusan kita sebelumnya tentang autentikasi?"

"Cari di basis kode kita bagaimana kita menangani pembatasan laju API"

"Tunjukkan konteks terkait tentang sistem pembayaran"

Contoh pelajaran

AI harus secara otomatis menangkap pelajaran saat Anda mengungkapkan frustrasi atau koreksi:

Pengguna berkata

"TIDAK! Kamu push tanpa menjalankan tes dan sekarang produksi rusak!"

→ AI menangkap pelajaran dengan keparahan: kritis, kategori: verifikasi

Pengguna berkata

"Itu salah. Selalu gunakan snake_case untuk kolom basis data, bukan camelCase."

→ AI menangkap pelajaran dengan keparahan: sedang, kategori: kualitas_kode

Pengguna berkata

"Saya lebih suka mode ketat TypeScript. Tolong selalu aktifkan."

→ AI menangkap pelajaran dengan keparahan: rendah, kategori: spesifik_proyek

Pelajaran ini secara otomatis dimunculkan di sesi mendatang saat konteks yang relevan diminta.

Pemeliharaan

Tetap terkini.

Untuk mendapatkan fitur terbaru, perbaikan bug, dan peningkatan, perbarui server MCP secara berkala:

terminal · terminal · macOS / Linux

curl -fsSL https://contextstream.io/scripts/mcp.sh | bash

terminal · powershell · Windows

irm https://contextstream.io/scripts/mcp.ps1 | iex

Server MCP akan secara otomatis memperingatkan Anda ketika versi yang lebih baru tersedia. Setelah memperbarui, mulai ulang alat AI Anda untuk menggunakan versi baru.

Pemecahan masalah

Ketika sesuatu tidak berfungsi.

Server MCP tidak mulai

Pastikan contextstream-mcp terinstal dan tersedia di PATH Anda. Coba jalankan contextstream-mcp --version secara manual untuk memeriksa kesalahan.

Kesalahan autentikasi

Verifikasi kunci API Anda benar dan belum kedaluwarsa. Anda dapat membuat kunci baru dari dasbor ContextStream Anda.

Alat tidak muncul

Mulai ulang aplikasi AI Anda setelah memodifikasi konfigurasi. Periksa log aplikasi untuk kesalahan koneksi MCP.

Tidak ada ruang kerja ditemukan (penyiapan pertama kali)

Jika akun Anda belum memiliki ruang kerja, ContextStream akan meminta asisten AI Anda untuk menanyakan nama ruang kerja. Folder saat ini akan dibuat sebagai proyek. Lihat Alat MCP untuk workspace_bootstrap.

Langkah selanjutnya

Lanjutkan menjelajah.

Penyiapan TimUndang anggota, bagikan konteks.Baca Pelajaran yang DipetikTangkap kesalahan, jangan pernah mengulanginya.Baca Peristiwa MemoriPelajari tentang tipe memori.Baca Pencarian SemantikCari berdasarkan makna.Baca

{"@context":"https://schema.org","@type":"Organization","name":"ContextStream","url":"https://contextstream.io","logo":"https://contextstream.io/logo.png","description":"Give your AI applications infinite context. Unify code, documentation, and workspace history into a single, queryable intelligence layer with semantic search and knowledge graphs.","sameAs":["https://twitter.com/contextstream","https://github.com/contextstream"]}
{"@context":"https://schema.org","@type":"SoftwareApplication","name":"ContextStream","applicationCategory":"DeveloperApplication","operatingSystem":"Any","description":"Give your AI applications infinite context. Unify code, documentation, and workspace history into a single, queryable intelligence layer with semantic search and knowledge graphs.","offers":{"@type":"Offer","price":"20","priceCurrency":"USD","description":"Pro plan includes a 5-day free trial"},"aggregateRating":{"@type":"AggregateRating","ratingValue":"5","ratingCount":"10"}}