local-brain-mcp
Enables AI coding assistants to recall past lessons, bugs, and conventions from local git history with sub-5ms latency and full privacy.
README
š§ local-brain-mcp
Local-first, zero-latency, git-aware AI memory for Claude Code, Cursor, Copilot & Windsurf.
No cloud. No API keys. No privacy risk. Sub-5ms recall.
Why local-brain?
| Problem with cloud AI memory tools | How local-brain solves it |
|---|---|
| š 150ā800ms network latency per recall | ā” < 5ms ā local SQLite vector search |
| āļø Your code sent to foreign servers | š 100% on-device, zero egress |
| šø Token bloat on every session | š¦ Hard 250-token budget cap per recall |
| šļø Stale outdated context from old decisions | š Git-diff invalidation marks old memories STALE |
| š WIP/typo commits pollute the brain | šÆ Conventional commit filter keeps only high-signal lessons |
| šļø Monorepo noise across packages | šÆ Path-scoped queries, per-package namespacing |
Quick Start
# 1. Install
npm install -g local-brain-mcp
# 2. Init ā auto-detects Claude Code, Cursor, Copilot, Windsurf
npx local-brain init
# 3. Ingest your git history
cd /path/to/your-project
npx local-brain ingest
# 4. Restart your AI editor ā the MCP tools are now available
MCP Tools
brain_recall
Semantic search your codebase memory. Results capped to 250 tokens.
{
"query": "JWT auth bug",
"file_path": "src/auth/jwt.ts",
"max_items": 5,
"category": "fix"
}
Example output:
## Brain Recall: "JWT auth bug"
⢠[src/auth/jwt.ts @ 8a4f12] (fix): JWT refresh race condition ā RS256 cert rotates every 24h. Never use HS256 in dev.
⢠[src/auth/session.ts @ c31d04] (bug): Sessions expire silently on Tuesdays 02:00 UTC ā auth service maintenance window.
brain_learn
Manually store a lesson or team convention.
{
"lesson": "Always seed the test DB before running Playwright tests or auth flows break.",
"category": "convention",
"file_path": "tests/setup.ts"
}
brain_trace
Full history of all memories for a specific file.
{
"file_path": "src/db/client.ts"
}
brain_prune
Remove stale/deprecated memories. Optionally run the full git invalidation pass.
{
"status": "stale",
"run_invalidation": true
}
CLI Commands
local-brain init # setup wizard ā writes MCP config for all detected editors
local-brain ingest # scan git history and build the brain DB
local-brain ingest --since "6 months ago" --verbose
local-brain status # show DB memory counts
local-brain prune --invalidate # detect + remove stale memories
How It Works
git history
ā
[git-ingest.ts] ā filters WIP/typo/format commits
ā
[embeddings.ts] ā pure-JS TF-IDF feature hashing (sub-1ms, zero network)
ā
[db.ts] ā stores in .git/brain.db (Float32Array BLOB vectors in SQLite)
ā (on file change)
[invalidation.ts] ā marks stale if file changed > 30%
ā (on MCP tool call)
[recall.ts] ā in-memory cosine similarity search, scoped to package, capped to 250 tokens
ā
Claude Code / Cursor / Copilot / Windsurf / Zed
Supported Editors
| Editor | Config auto-detected |
|---|---|
| Claude Code | ~/.claude.json |
| Cursor | ~/.cursor/mcp.json |
| Windsurf | ~/.codeium/windsurf/mcp_config.json |
| VS Code Copilot | ~/.vscode/mcp.json |
| Zed | ~/.config/zed/settings.json |
Tech Stack
- Protocol:
@modelcontextprotocol/sdk(StdioServerTransport) - Storage:
better-sqlite3(SQLite WAL mode with BLOB vectors) - Embeddings: Pure-JS TF-IDF feature hashing (384-dim, sub-1ms, 100% offline)
- Git Engine:
simple-git - CLI Engine:
commander
License
MIT ā build freely.
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.