Mneme
Give Claude Code and other MCP clients vault-native, accountable long-term memory where markdown files remain the source of truth.
Overview
Mneme is an MCP server for the build phase that persists accountable agent memory in a markdown vault for Claude Code and other MCP clients.
What is this MCP server?
- Vault-native memory: markdown on disk is ground truth, not opaque vector-only stores
- Accountable memory model aimed at traceable updates for coding agents
- stdio transport; npm package mneme-mcp-server at version 2.0.2
- runtimeHint npx for quick local registration without a global install
- Targets Claude Code and generic MCP clients
- Server package version 2.0.2
- 1 npm package (stdio transport)
- runtimeHint: npx
What problem does it solve?
Coding agents forget context between sessions unless you paste the same README fragments again, and opaque memory tools hide what was stored.
Who is it for?
Solo builders using Claude Code who want Git-friendly, markdown-first memory across long-running products.
Skip if: Builders who only need a single-session chat with no persistence, or teams that require a hosted multi-user memory SaaS with no local vault.
What do I get? / Deliverables
After registering the stdio server via npx, your agent can read and update vault markdown memory you own, version, and audit in your repo.
- MCP-backed read/write access to vault markdown memory from the agent
- Session-to-session continuity for architecture and decision notes
- Human-auditable memory artifacts in your filesystem
Recommended MCP Servers
Journey fit
Persistent agent memory is an build-time capability you add when wiring how your coding agent remembers projects across sessions. Mneme ships as an npm stdio MCP server run via npx, which is classic agent-tooling configuration alongside Claude Code.
How it compares
Local vault memory MCP integration, not a cloud observability or database-admin MCP server.
Common Questions / FAQ
Who is Mneme for?
Indie developers and Claude Code users who want durable, file-based agent memory they can inspect in markdown and track in version control.
When should I use Mneme?
Use it during build when a project spans many agent sessions and you need consistent facts, decisions, and notes without re-prompting from scratch.
How do I add Mneme to my agent?
Add an MCP stdio server entry with runtimeHint npx and identifier Mneme-mcp-server (v2.0.2), then point it at your vault directory per the GitHub repo setup.