Marm Mcp Server
Give your coding agent durable memory and semantic recall across sessions so context survives beyond a single chat window.
Overview
io.github.Lyellr88/marm-mcp-server is a MCP server for the Build phase that provides advanced AI memory capabilities and semantic search for coding agents.
What is this MCP server?
- Universal MCP server with advanced AI memory and semantic search
- Published as marm-mcp-server on PyPI and OCI image lyellr88/marm-mcp-server:2.9.2
- Stdio transport for standard MCP clients
- Version 2.9.2 in server manifest; source under Lyellr88/MARM-Systems
- Cross-session recall for project facts, decisions, and embeddings-backed lookup
- MCP server version 2.9.2
- Distribution: PyPI marm-mcp-server and OCI lyellr88/marm-mcp-server:2.9.2
- Repository github.com/Lyellr88/MARM-Systems
Community signal: 303 GitHub stars.
What problem does it solve?
Your agent forgets project context every new session and keyword search in notes fails when you only remember concepts, not exact phrases.
Who is it for?
Solo builders running long agent-assisted projects who need local or self-hosted memory with MCP-native read/write tools.
Skip if: Teams that require a vendor-managed enterprise memory SaaS with compliance attestations out of the box, or one-off scripts with no cross-session needs.
What do I get? / Deliverables
After registration, the agent can persist and semantically retrieve memories so multi-day solo builds stay coherent with less repeated explanation.
- Running MARM MCP with memory write and semantic query tools
- Cross-session recall of project decisions and notes via the agent
- Embeddings-backed search without manual prompt stuffing
Recommended MCP Servers
Journey fit
Persistent memory and semantic search are foundational agent infrastructure while you build products with long-horizon AI assistance. MARM exposes memory storage and retrieval primitives—classic agent-tooling that complements codegen, debugging, and planning MCPs.
How it compares
Self-hosted memory and semantic-search MCP, not a skill marketplace router or NinjaONE-style monitoring bridge.
Common Questions / FAQ
Who is io.github.Lyellr88/marm-mcp-server for?
It is for developers who want a universal MCP memory layer with semantic search for Claude Code, Cursor, Codex, or compatible stdio clients.
When should I use io.github.Lyellr88/marm-mcp-server?
Use it during Build (and ongoing iteration) when session length and repeated context make ad-hoc chat history insufficient.
How do I add io.github.Lyellr88/marm-mcp-server to my agent?
Install marm-mcp-server from PyPI or pull the OCI image lyellr88/marm-mcp-server:2.9.2, configure stdio in your MCP settings per MARM-Systems docs, then restart the agent.