Command Palette

Search for a command to run...

Home / Servers

ExternalBrain

Updated 14d ago

by bejranonda

External Brain: Self-Improving, Self-Hosted AI Coding Memory Across Every Tool, Project, and Team

Your AI's memory is trapped in one tool, one project, one person. Claude Code doesn't share with Cursor, a lesson learned on one repo doesn't carry to the next, and your teammates each start from zero. External Brain is one shared, self-hosted knowledge layer across every MCP tool, project, and team. A skill learned once applies everywhere it's relevant, you can inspect and edit what it knows, and answers cite your own work. Best of all, it improves on its own: after each session it proposes new skills from what you did, so every project gets sharper without anyone hand-writing rules. Built for teams and enterprise, on your own infrastructure.

image

GitHub Workflow CI build status GitHub Stars count GitHub Forks count GitHub Open Issues MIT License TypeScript strict mode Model Context Protocol native support Self-hosted via Docker Compose

External Brain is a self-hosted MCP (Model Context Protocol) server + webapp that gives AI coding agents persistent, long-term memory. It ingests your coding sessions, extracts durable knowledge (skills, rules, recipes, anti-patterns), retrieves it by semantic meaning when you start a new task, and answers questions about your own codebase through a grounded Oracle — every answer cited back to the sessions and skills that support it.

Unlike the memory built into each AI tool, that store is one shared, inspectable layer across every MCP client — it stays on your own infrastructure instead of sitting in a separate black box locked inside each tool.

Provider-agnostic (Google Gemini, GLM, OpenAI, Anthropic Claude), runs on a single VM with Docker Compose, and MIT-licensed — fork it and build your own.

Architecture diagram showing how External Brain connects AI coding tools like Claude Code and Cursor via MCP to a Postgres + pgvector knowledge store


Why Use External Brain? — One Memory Across Every Tool, Project, and Team

Modern AI coding tools have memory now. The real problem is where that memory lives and how far it reaches:

  • Siloed — per tool, per project, per person. Claude Code doesn't share with Cursor or Copilot; a lesson learned on one repo doesn't carry to the next; and your teammates each start from zero. Knowledge that should compound stays stuck in one place.
  • A black box. You can't see what it kept, fix it when it's wrong, or curate it. You just hope it remembered the right thing.
  • Not yours. It's locked inside one vendor's cloud, tied to that one tool. You can't inspect it, move it, or share it on your terms.

Related servers

n8n

Updated today

by n8n-io

Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.

204,501

mcp-server-fetch

OfficialUpdated 14d ago

by modelcontextprotocol

A Model Context Protocol server providing tools to fetch and convert web content for usage by LLMs

90,371

@modelcontextprotocol/server-everything

OfficialUpdated 14d ago

by modelcontextprotocol

MCP server that exercises all the features of the MCP protocol

90,371

mcp-server-git

OfficialUpdated 14d ago

by modelcontextprotocol

A Model Context Protocol server providing tools to read, search, and manipulate Git repositories programmatically via LLMs

90,371