yoheinakajima

yoheinakajima / selfMCP

Public

A self-extending MCP server

19
2
100% credibility
Found Apr 12, 2026 at 19 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

selfMCP is a persistent skill library that enables AI assistants to search, create, update, and execute custom abilities stored in a simple local file.

How It Works

1
🔍 Discover selfMCP

You hear about selfMCP, a magic library that lets your AI assistant learn and create new tricks on its own.

2
Pick your setup
☁️
Online playground

Copy to a free site like Replit, press run, and it's live instantly.

💻
Your computer

Download and start it with simple steps to keep everything local.

3
🔗 Connect your AI

Link it to your favorite AI chat app, like Claude, so they can talk and share skills.

4
📋 Browse ready skills

Your AI sees a list of starter abilities, like checking what tools are available.

5
✨ AI builds new ones

When your AI needs a new trick, it creates and saves it automatically for next time.

6
▶️ Run a skill

Tell your AI to use a skill with your own details, and watch it work.

🎉 AI grows smarter

Over time, your assistant collects more skills and handles tougher tasks effortlessly.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 19 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is selfMCP?

selfMCP is a Python MCP server that creates a self-extending toolkit for LLMs like Claude Desktop, Claude.ai, or Cursor. It provides a skill registry with hybrid search, execution runtime, and CRUD tools—all in a single SQLite file—so models can discover, run, create, or update Python skills on the fly. Skills inherit your server environment, pulling in API keys seamlessly without extra plumbing.

Why is it gaining traction?

Its bootstrap-only design lets LLMs bootstrap everything else, turning one-time setup into a growing agent toolbox. Standouts include pre-execution auth checks with setup hints, vector+keyword search across skills, and dead-simple Replit deploys via streamable-HTTP at /mcp. No vector DB ops or external services—just env vars for embeddings and skills that evolve across sessions.

Who should use this?

MCP experimenters building persistent LLM agents in Claude or Cursor who hate rebuilding tools from scratch. Replit hackers prototyping AI workflows with API integrations like Anthropic or OpenAI. Python devs wanting a local stdio server for self-improving scripts without Docker hassle.

Verdict

Grab it if you're into self-extending MCP servers—docs are thorough, tests cover core flows, and 19 stars show early buzz. But 1.0% credibility flags its raw maturity; production users should monitor for scale.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.