https://github.com/LastEld/AMS
AMS – Deterministic Agent Pipeline with A2A‑style Orchestration and Cryptographic Audit
AMS v5.0.0
AMS Documentation
AMS v5.0.0 is an audit-first operating layer for MCP clients. It gives Claude Code, Codex, Obsidian, and other assistants durable scope, task state, proof-oriented execution trails, long-term memory, and orchestration primitives.
This site is organized around the live product surface and the operating model, not around stale file names or old release narratives.
Version: 5.0.0 (Obsidian Control Center & Sub-Agent Architecture)
Repository: LastEld/AMS
Server entrypoint: E:/AMS/src/server.js
Runtime verifiedAMS is currently operational and readiness is green.
Tool surface484 visible tools in the full profile, with a narrower PM/operator lane.
Project scale243 total tasks, 150 completed, 5 roadmaps.
Desktop flowObsidian sync, diagnostics, and managed notes are documented with live evidence.
Project Status
See the full Project Status Dashboard for:
- what is completed
- what remains active
- current next actions
- live runtime metrics
Quick stats:
- | 243 total tasks | 150 completed (62%) |
| ------------------- | ----------------------- | - | 0 active watchers during PM verification | 5 roadmaps |
| -------------------------------------------- | -------------- | - 484 visible tools in the full profile
For the proof-backed desktop story, read Live evidence.
Choose a Path
Install and verify
Start with Getting started, then wire your client in Client setup. If you use Obsidian, the Obsidian plugin guide covers the vault projection flow, the automatic note tree, and the split between TODOS.md, PROGRESS.md, and Dashboard.md.
Operate day to day
Read the User guide, then jump to Common workflows.
Understand the system
Use the Architecture overview for runtime, storage, proof, and reliability design, then jump to Dependency graphs for module boundaries.
Run it in an environment
Go from Operations to the Configuration guide, the Distribution guide, and then the Deployment guide.
Architecture At A Glance
Key model: AMS is not a stateless helper around a chat client. It is a durable control plane that keeps task state, audit history, thought trails, and memory coherent across multiple clients.
Release Highlights
Obsidian Control Center
- sync settings, diagnostics, and queue controls in the desktop plugin
- managed vault notes for active work, history, and dashboard views
- proof-backed operator screenshots published in the docs
- corrected task pagination so progress notes can render the full task set instead of only the first
task_listpage
PM / Executor Pipeline
- Codex as PM and reviewer
- Claude as executor and verifier
- Obsidian as a projection layer, not a second database
- GSD workflows and agent pools as the reusable execution structure
Proof and Runtime Stability
- audit, thought, and Merkle flows remain the proof chain
- desktop verification uses the runtime-first bootstrap sequence
- current PM verification window found 0 log matches for
SQLITE_BUSY,database is locked, andTransport closed
What AMS Provides
- persistent project, task, roadmap, and workflow state
- audit sessions, thought trails, and Merkle verification
- retrieval-backed memory and context packaging
- GSD-style orchestration for planned execution
- a large full-profile MCP surface with a narrower PM/operator path
- an Obsidian plugin that exposes a note-facing desktop control surface
Core Runtime Sequence
unified_init
unified_vitals
unified_help
unified_set_project
Use runtime discovery before relying on static documentation for exact tool inventory.
Interactive desktop clients should default to AMS_WATCH_MODE=none. Use watcher-backed sync only on a dedicated AMS runtime that explicitly owns that responsibility.
Public Documentation Map
- Project Status
- Tutorials
- Obsidian plugin
- Architecture
- Dependency graphs
- API guidance
- Tool Reference — all 484 tools
- Reference
- Live evidence
- Operations
- Configuration
- Distribution
- Active surface
Documentation Structure
tutorials/: install, connect, verify, and operate AMSarchitecture/: runtime, storage, proof, reliability design, and dependency graphsapi/: runtime-first guidance for the tool surfacereference/: durable reference material and live evidenceoperations/: deployment, environment, and operational checksexamples/: worked workflows and transcriptsarchive/: historical reports and retired narrativesinternal/: planning artifacts retained for maintainers
Release: v5.0.0 on GitHub
License: Apache-2.0 WITH Commons-Clause
This site is open source. Improve this page.
Máy chủ liên quan
Scout Monitoring MCP
nhà tài trợPut performance and error data directly in the hands of your AI assistant.
Alpha Vantage MCP Server
nhà tài trợAccess financial market data: realtime & historical stock, ETF, options, forex, crypto, commodities, fundamentals, technical indicators, & more
Serencp
VM serial console viewer
Model Context Protocol servers
A collection of reference implementations for the Model Context Protocol (MCP), showcasing servers implemented with TypeScript and Python SDKs.
TCC
Automatically generates MCP servers from OpenAPI specifications, enabling conversational AI agents to interact with existing web systems.
Diffchunk
Navigate large diff files with intelligent chunking and navigation tools.
Next.js DevTools MCP
next-devtools-mcp is a MCP server that provides Next.js development tools and utilities for AI coding assistants like Claude and Cursor.
Remote MCP Server (Authless)
An example of a remote MCP server deployable on Cloudflare Workers without authentication.
Packmind
Access and manage your team's coding best practices and knowledge base from Packmind.
Code Summarizer
A command-line tool that summarizes code files in a directory using Gemini Flash 2.0.
ContextKeeper
Provides perfect memory for AI-assisted development by capturing project context snapshots, enabling natural language search, evolution tracking, and code intelligence.
refactor-mcp
Refactor code using regex-based search and replace.