MarsNMe
Give Claude Code or Cursor a Supabase-backed memory layer so project context survives across chats and profiles.
Overview
MarsNMe is a MCP server for the Build phase that stores and retrieves profile-scoped agent memory via Supabase and embedding search.
What is this MCP server?
- Agent-agnostic memory backend with MCP_PROFILE separation for multiple memory namespaces
- npm package @marsnme/mcp-gateway (v0.2.0) with streamable-http on localhost (default ports per profile, e.g. coco=18790)
- Supabase REST storage plus Jina embeddings for semantic recall tools
- Required env: MCP_PROFILE, SUPABASE_BASE_URL, SUPABASE_SERVICE_ROLE_KEY, JINA_API_KEY
- Designed for human–AI continuity over time, not single-session context windows
- Server version 0.2.0
- 4 required environment variables (profile, Supabase URL, service role, Jina key)
Community signal: 6 GitHub stars.
What problem does it solve?
Every new agent session forgets what you already decided, so you repeat explanations and lose thread on long-running solo builds.
Who is it for?
Indie builders running Claude Code or Cursor who want externalized, searchable memory tied to Supabase—not just CLAUDE.md notes.
Skip if: Teams that only need a static AGENTS.md in-repo or cannot host Supabase credentials and a Jina API key.
What do I get? / Deliverables
After registration, your agent can persist and query memories under a named profile so continuity carries across sessions and tools.
- Local or hosted MCP endpoint for memory read/write tools
- Isolated memory namespaces per MCP_PROFILE
- Embedding-backed recall over stored agent memories
Recommended MCP Servers
Journey fit
Persistent agent memory is infrastructure you wire in while assembling your coding agent stack, before you rely on the agent for day-to-day product work. MarsNMe exposes MCP tools for profile-scoped recall and embedding search—classic agent-tooling, not app feature code.
How it compares
Long-term memory MCP backend, not a one-shot RAG document indexer or chat UI feature.
Common Questions / FAQ
Who is MarsNMe for?
Solo and small-team builders who use multiple AI agents and need separated, durable memory profiles per product or client.
When should I use MarsNMe?
When you are wiring agent-tooling for a real project and session resets are costing you rework on architecture and task state.
How do I add MarsNMe to my agent?
Install @marsnme/mcp-gateway via npx, set MCP_PROFILE and Supabase/Jina env vars, expose the streamable-http MCP URL (default local port by profile), and register that endpoint in Claude Code, Cursor, or another MCP client.