Gitmem
Give your coding agent persistent institutional memory across sessions so decisions, conventions, and context compound instead of resetting every chat.
Overview
gitmem is a MCP server for the Build phase that provides persistent institutional memory for AI coding agents so context compounds across sessions.
What is this MCP server?
- npm stdio package gitmem-mcp (registry version 1.2.1)
- Persistent institutional memory designed to compound across agent sessions
- Free tier without Supabase; optional SUPABASE_URL and SUPABASE_SERVICE_ROLE_KEY for pro semantic search and analytics
- stdio transport for local MCP clients
- Positioned explicitly for AI coding agents, not generic note apps
- Server schema version 1.2.1
- One npm stdio package identifier: gitmem-mcp at version 1.2.1
- Two optional environment variables for pro tier: SUPABASE_URL, SUPABASE_SERVICE_ROLE_KEY
Community signal: 8 GitHub stars.
What problem does it solve?
Your agent forgets architecture decisions and project quirks every session, forcing you to re-explain the same institutional knowledge.
Who is it for?
Solo builders running long-lived agent workflows who want a dedicated memory MCP beyond ephemeral chat threads.
Skip if: One-off scripts or builders who only need GitHub or ticket integrations without cross-session knowledge retention.
What do I get? / Deliverables
After installing gitmem-mcp via stdio, your agent can read and write durable memory so later tasks reuse accumulated project context without repeating setup lore.
- stdio MCP connection to gitmem-mcp for read/write institutional memory
- Cross-session context retention for agent-assisted development
- Optional Supabase-backed semantic search when pro env vars are configured
Recommended MCP Servers
Journey fit
Long-horizon agent memory is infrastructure for how you build with AI assistants, so the canonical shelf is Build → agent-tooling even though recall helps Operate and Grow documentation. gitmem extends the agent runtime with durable memory stores—not app frontend, backend APIs, or Git host integrations—fitting agent-tooling precisely.
How it compares
Agent memory MCP via npm stdio, not a GitHub API bridge or a PR risk-screening service.
Common Questions / FAQ
Who is gitmem for?
It is for indie developers using AI coding agents who need institutional memory that persists and compounds across conversations and repos.
When should I use gitmem?
Use it during Build (and ongoing Operate) when you want the agent to recall prior decisions, conventions, and facts without reloading full README and chat exports each time.
How do I add gitmem to my agent?
Add the npm MCP package gitmem-mcp with stdio transport in your MCP client; set SUPABASE_URL and SUPABASE_SERVICE_ROLE_KEY only if you enable pro-tier semantic search and analytics.