Agent Memory Mcp
Give coding agents durable, queryable memory with provenance and decay-weighted recall instead of stuffing everything into the chat context.
Overview
Agent Memory MCP is a MCP server for the build phase that persists agent memories in SQLite with provenance, decay-weighted recall, and feedback loops.
What is this MCP server?
- SQLite-backed memory store with configurable MEMORY_DB_PATH (default ~/.agent-memory/memory.db)
- Provenance tracking so recalled facts can be traced to source interactions
- Decay-weighted recall to prioritize fresher or reinforced memories over stale notes
- Feedback loops to improve what the agent retrieves over time
- Stdio npm package @kiraautonoma/agent-memory-mcp version 0.1.3
- Package version 0.1.3
- Default database path ~/.agent-memory/memory.db
- Transport: stdio npm registry
What problem does it solve?
Agents forget prior decisions across sessions and you cannot trust unstructured chat history as a durable knowledge base.
Who is it for?
Indie builders running multi-session coding agents who want local, structured memory without a separate vector SaaS on day one.
Skip if: Teams that need enterprise RBAC, multi-user shared knowledge graphs, or compliance-reviewed memory governance out of the box.
What do I get? / Deliverables
After install, your agent can store and retrieve weighted memories from a local database with traceable provenance.
- Local SQLite memory database for agent recall
- MCP-accessible memory operations with provenance and decay-weighted retrieval
Recommended MCP Servers
Journey fit
Persistent agent memory is added while you integrate MCP tooling into the product you are building, which is the build journey phase. Memory servers sit on the agent-tooling subphase shelf next to other MCP backends that extend what Claude Code and similar hosts can remember across sessions.
How it compares
Durable agent memory MCP backend, not a one-shot prompt skill or generic note-taking plugin.
Common Questions / FAQ
Who is Agent Memory MCP for?
It is for solo builders and agent authors who want persistent, recall-ranked memory exposed as MCP tools to coding agents.
When should I use Agent Memory MCP?
Use it during build when your agent repeats work because context windows drop project facts, preferences, or prior debugging outcomes.
How do I add Agent Memory MCP to my agent?
Install @kiraautonoma/agent-memory-mcp via npm stdio transport, set MEMORY_DB_PATH if you want a custom SQLite file, and register the server in your MCP host config.