Bring the context.
Choose your sources, add guidance, review the proposed memories, and approve the import.
A shared memory layer for the work your AI needs to carry forward. Decisions, sources, and corrections stay connected across the tools you use.
Connected through MCP, REST, and supported clients.
Useful memory is more than a matching passage. It needs the decision, the source behind it, and a way to handle what changes.
The 2.0 server implementation adds an Associative Memory Graph: related memories can activate together, use can reinforce connections, and consolidation can connect detailed memories to higher-level context.
Read the implementation and availability notes →Follow a decision, its related context, and a later correction. Evidence-aware ranking and Veritas lineage help keep those distinctions visible.
How memory works →Reconnect retries must be idempotent.
The decision stays linked to the meeting notes it came from.
Choose your sources, add guidance, review the proposed memories, and approve the import.
Organize approved material, extract facts and relationships, and preserve the source behind each memory.
Recall relevant context through a connected tool. Keep corrections, competing evidence, and history visible.
Loom keeps source review and approval before the processing handoff.
Choose sources → review candidates → approve the handoff. Use the dashboard to run a real import with your own material.
Set up your first connection →Connect an MCP client, use a supported extension, or build with the REST tools API. Recall and save are explicit actions; connection methods vary by client.
Start with a free account. Explore team plans or licensed self-hosting as your work grows.