# Mnemo

Persistent memory for Claude Code that captures decisions, conventions and preferences during a session and recalls them later by meaning, with sub-100-millisecond semantic search, instead of relying on a single static CLAUDE.md file.

Mnemo installs a CLI, an MCP server and a Claude Code skill, after which Claude uses the tool automatically when relevant; a post-edit hook can also auto-capture any change to files such as CLAUDE.md, AGENTS.md or files under docs/decisions and docs/adr, turning them into project memories without manual steps. A knowledge-graph layer lets entities such as services be created, linked to each other, and queried for what depends on them alongside their attached memories.

Recall is scored as a weighted combination of cosine similarity to the query, how recently a memory was created, with a 30-day half-life decay, and how often it has been accessed, saturating at 20 reads, so relevant and recent memories surface ahead of older, rarely used ones. It supports cross-agent use, plugins and syncing to a self-run sync backend that the project states never sees plaintext, and can encrypt its local memory database at rest with a passphrase.

It runs entirely locally with no daemon, cloud account or telemetry, installs from npm, and is written in TypeScript under the MIT license.

Website: https://github.com/omermaksutii/mnemo