Stop re-explaining your project to your AI every session. memtomem turns your notes, docs, and code into a searchable memory that any MCP-compatible agent can use — across sessions, across agents, all on your machine.
Check Python and uv, install memtomem, then initialize the no-download Minimal preset.
Run mm status, mm add, and mm search. A source-backed result is the success gate.
Connect one AI client, index existing content, verify a new session, or add STM to an MCP server.
All context is lost when a session ends. Architecture decisions, coding patterns, and debugging history must be re-explained every time.
Knowledge from Claude Code can't be carried over to Cursor. Each agent is trapped in its own isolated memory silo.
Current memory systems only work when agents explicitly search, are locked to specific runtimes, and offer only a single LTM layer.
Index your notes, docs, and code with mm index, then find them with hybrid search — BM25 keyword and dense-vector semantic search fused via RRF, so exact identifiers and meaning-based queries both land. Markdown, code, and structured files are chunked by structure, and re-indexing only re-embeds the chunks that changed.
Memory doesn't vanish when a session ends. Namespaces route searches between agent-specific and shared scopes; they are not access-control boundaries. The session workflow lets Claude Code, Cursor, and Codex continue from one memory store.
Keep master skills, commands, and subagents in the .memtomem/ Store. Preview and Pull runtime copies into the Store, then Push the reviewed Store version to selected runtimes. Copy or move artifacts across projects and tiers, or operate across registered projects in one batch.
Your agent doesn't have to ask. STM observes the MCP calls it proxies and surfaces relevant memories at the right moment. Each surfaced memory carries an id, so your agent can rate or invalidate individual items.
Responses from MCP tools routed through STM can be compressed before they reach your agent. When a response exceeds the context budget, one of 10 strategies is selected by content type. Actual savings depend on the response and configured retention limits.
SQLite storage and local ONNX inference need no GPU or cloud account. Optional Ollama, OpenAI-compatible, Cohere, and Langfuse integrations can contact configured services. Local files use restrictive permissions, and credential-shaped content is filtered at cache, index, and share boundaries.
Install LTM and verify a durable memory round trip before connecting a client.
Getting StartedConnect Claude Code, Codex, or another released MCP client and verify mem_status.
Index notes and code or import built-in AI memory with a preview and source check.
ContentSave a confirmed decision, start a new session, and retrieve its reason and source.
WorkflowRun the local demo, make a real proxied call, inspect metrics, and restore the original path.
Optional STMReturn to the failed install, search, MCP, Web UI, or STM step with a concrete check.
RecoveryNo GPU is required. The default path stays local; external providers remain optional.