Muninn
Provides persistent memory for AI coding agents via MCP, allowing them to recall fragility, decisions, and bugs across sessions.
README
Muninn
Your AI coding agent forgets everything between sessions. Muninn fixes that.
npx muninn-ai
Muninn gives AI coding agents persistent memory via MCP. Before every edit, the agent knows what's fragile, what decisions were made, and what broke last time. After every session, it writes back what it learned. Knowledge compounds automatically.
Works with: Claude Code, Cursor, Windsurf, Continue.dev, and any MCP-compatible tool.
The Problem
AI coding agents are stateless. Every session starts from zero:
- They break things — editing a critical file with no idea it's fragile
- They contradict themselves — picking a different pattern than 3 sessions ago
- They repeat mistakes — re-introducing bugs that were already fixed
- They lose context — forgetting why something was done a certain way
4 Tools
Muninn exposes 4 tools to your AI agent:
| Tool | Purpose |
|---|---|
recall |
Pre-edit context — fragility, co-changers, related decisions, open issues, blast radius |
remember |
Record decisions and learnings — auto-categorized, searchable across sessions |
track |
Bug lifecycle — add when found, resolve when fixed, surface when relevant |
muninn |
Everything else — status, reindex, fragile files, decision outcomes |
What Compounds
Muninn gets smarter with every session:
- Fragility scoring — weighted composite of dependents, test coverage, change velocity, error history, complexity, and export surface
- File correlations — tracks which files change together to predict co-changes and warn about missed updates
- Learning graduation — patterns that prove true get promoted; contradicted ones get archived
- Decision grounding — architectural choices link to outcomes so the agent knows what worked and what didn't
Quick Start
Option 1: npx (Recommended)
npx muninn-ai
Installs Bun (if needed), clones the repo, registers the MCP server, and sets up hooks. One command.
Option 2: Manual Install
git clone https://github.com/ravnltd/muninn.git ~/.local/share/muninn
cd ~/.local/share/muninn && ./install.sh
First Run
On first session, Muninn auto-indexes your git history (last 100 commits) so recall returns useful context immediately. No setup needed.
Multi-Editor Support
After install, Muninn auto-detects and configures:
- Claude Code — MCP server + session hooks
- Cursor —
.cursor/mcp.json - Windsurf —
.windsurf/mcp.json - Continue.dev —
.continue/config.json
Or register manually:
muninn setup --list # See what's detected
muninn setup --all # Configure all detected editors
Architecture
Bun + TypeScript + SQLite (via libsql/sqld). Runs as an MCP server over stdio. Local mode stores everything in .muninn/ per-project. HTTP mode connects to a shared sqld instance for multi-machine setups.
License
AGPL-3.0-only — free to use, modify, and share. If you run a modified version as a network service, you must release your source.
Built in collaboration with Claude Code.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.