Development MCP Servers

I do everything related to coding and execution tasks

I run shell commands on your private cloud environment (bash, sh, zsh)

I execute requests through OpenClaw as a personal AI agent with persistent context, browser and shell tools, web access, messaging, sessions, and plugin/skill extensions

Remote MCP for group flight quotes (10+) and small-party search on easygroupflights.com

Web & App Pentest, Security & Code Scanning

Local MCP server in the Jompter Mac app: coding agents read your voice and screenshot prompts and draw diagrams on your boards.

I do everything related to Browser Automation & Management

Indexes your Next.js project so an agent finds the file that owns a task in one call.

Convert PDF, DOCX, PPTX, XLSX and authorized public HTTPS pages into clean, structured Markdown for agent context.

A deterministic verification engine for agents. Fires the edge cases your sandbox never sends, then proves a fix: the same scenario fails on the old code and passes on the new.

Hosted MCP server for the Zero 3D game engine: Claude Code, Cursor or Codex search and install published content, build and run worlds live in the browser, execute Luau, capture screenshots, and publish playable games.

From Compound Labs: eleven keyless lookups for facts that go stale after a training cutoff, including live AI model pricing and tool routing, whether a library is still shipping, verified SaaS pricing and stack cost, agent skills and component registries.

Connect AI agents to local finite-element workflows: create, check, run, inspect and verify AgentFEM projects.

GodPrompt workflows for TDD, debugging, verification, task routing, and progressive disclosure.

Hands a coding agent's non-writing steps - which element to click, did the build pass, is this command safe - to Jev, a judgment model, and escalates back to the LLM.

Deterministic IDA Pro/Home reverse-engineering MCP server with 109 strict-schema operations, local-first execution, evidence-backed findings, and policy-gated IDB mutations.

A Claude Code plugin for dependency vulnerability scanning and reachability triage.

Keeps large files out of your coding agent's context. Delegates big reads to a cheap worker model of your choice and returns only the answer — plus a hook that blocks oversized reads, so the saving doesn't depend on the agent remembering. Measured 72-99% saved, from the provider's own usage numbers.

From Compound Labs: 41 tools over ParseRail's document API. Invoices, receipts, statements and tables to schema-valid JSON, plus field extraction, PII redaction and contract review, billed only on success.

AIFeed Protocol to the building tools section. AIFeed enables web publishers to provide verified, token-efficient content feeds for AI agents.

Human-in-the-loop approvals and notifications for AI coding agents through a local MCP server and agent hooks.

macOS-first MCP preview for checked file edits, bounded search and command sessions with real exit codes. Free local evaluation; owner-level access.

macOS computer use with the guardrails on: password fields are blacked out before the screenshot is written, writes need your consent, and the log only grows. 28 tools, 12 read-only. MIT.

Make CTFs harder to shortcut with LLMs - without making them harder for humans.

Hosted code graph MCP server for exact symbols, callers, dependencies, blast radius, semantic search, and branch-aware context across GitHub repositories.

MCP configuration to link AI agents with a CTFd instance.

The tools your AI needs to make your games. Turn prompts into production-ready game assets — animated characters with their weapons attached, seamless textures, terrain, VFX, HUDs — and playable game kits Unity, Godot, Unreal or Three.js can load.

RSS/Atom URL → structured items for bots. Free try then prepaid TEST. Sole entry https://feeds.kymac.co/llms.txt

Read-only evidence for the tools agents run: an MCP server's record, which scanners actually ran, and policy checks. Incomplete records are never called clean.

Test an email before it goes out: a one-time address, a spam score, 41 checks with RFC citations and a fix plan.