@memmo/cli
v0.8.3
Published
Long-term memory for AI coding agents — CLI + MCP server for Claude Code.
Downloads
700
Maintainers
Readme
memmo
Long-term memory for AI coding agents. This CLI connects a repository to Memmo and runs the MCP server + hooks that give Claude Code your team's approved memory before it acts.
Install
npm install -g @memmo/cliUsage
memmo login # once per machine; stores a token in ~/.memmo
memmo init # in a repo: registers it from the git remote, wires Claude Code, offers a scan
memmo status # check the setup
memmo desktop # wire Claude Desktop to this repo (init offers this when Desktop is installed)
memmo scan # propose starter memories (drives your local Claude Code)
memmo decision "…" # record a decision as memory
memmo sync # write approved memories into a managed CLAUDE.md blockNo per-repo config file: the repo is resolved from git remote get-url origin.
init writes .mcp.json (the server runs via npx -y @memmo/cli mcp, so teammates
need no global install), a CLAUDE.md section, and Claude Code hooks
(SessionStart · UserPromptSubmit · PreToolUse · SessionEnd).
MCP tools
get_repo_context()— the session briefing: rules, architecture, commands, riskssearch_memory(query)— approved memories, ranked by meaningget_relevant_warnings(files)— risks for the files about to changepropose_memories([...])— record durable knowledge (goes to the inbox);supersedescorrects an existing memoryrecord_session_summary(...)— submit what you did; Memmo proposes new memories
Point at a self-hosted backend with MEMMO_API_URL. For Claude Desktop (no
working directory) pass memmo mcp --repo <repoId>.
MIT licensed.
