Agent Memory MCP

agent-memory-mcp

Long-term memory and project-docs search for coding agents, in one local MCP server backed by SQLite.

Your agent remembers why a decision was made, finds the runbook that matches an incident, and notices when what it remembers no longer matches the repo.

Install

Claude Code. Nothing else to set up: no service, no model, one local database per project. Details.

/plugin install agent-memory --marketplace ipiton/agent-memory-mcp

Codex, Cursor, Gemini CLI, OpenCode, VS Code, Windsurf, or Claude Code without the plugin. A standalone server, which needs one embedding provider. Details, including builds without Homebrew.

brew tap ipiton/tap

brew install agent-memory-mcp cd your-repo agent-memory-mcp setup codex --yes

Restart the agent afterwards.

What it does

• Searches memory and docs together. Decisions, facts and procedures are ranked alongside your docs, ADRs and runbooks, and every hit says how far to trust it.
• Keeps memory clean. steward_run finds duplicates, contradictions and stale entries, and sends anything risky to a review inbox.
• Notices drift. drift_scan flags memories whose source file changed or disappeared.
• Leaves the last word to you. Notes become canonical only through promote_to_canonical. An agent cannot verify its own notes.
• Stays local. One binary and SQLite. An HTTP mode is there for a shared team instance.

> How many replicas should payments-api run, and why?

1. Why payments-api runs 4 replicas (relevance: 0.79) Trust: layer=raw source=decision confidence=0.96 freshness=1.00 owner=engineering Content: Decision 2026-09-12: payments-api runs 4 replicas, not 2. Two replicas dropped requests during the September card-network spike. Source: ./deploy/values.yaml.

More in Features.

Documentation

Development

make build

make test make vet

Conventions are in CONTRIBUTING.md. License: MIT.