Frequently asked questions
How Lore’s coordination memory layer works, how it compares, and what’s open source.
What is a coordination memory layer? +
A shared memory system for multi-agent AI: agents write events, the layer consolidates them into versioned facts, and serves budget-fit context packs with consistency guarantees — so every agent works from the same reality.
How is Lore different from Mem0? +
Mem0 optimizes single-user personalization with append-only memories ranked at read time. Lore targets team coordination: conflicts are resolved at write time under database constraints, and reads carry read-your-writes guarantees.
How is Lore different from Zep and Graphiti? +
Zep’s bi-temporal graph answers “what is true now.” Lore adds “will agent B see it in this pack” — sequence tokens, freshness lag, and a hot working-memory lane. Also: our OSS is a full server, not a library you operate around.
Does Lore work with LangGraph, CrewAI, AutoGen, or the Claude Agent SDK? +
Yes — SDKs for TypeScript and Python, and an MCP server for everything else. Lore is framework-neutral by design: no framework shares memory with a competitor’s agent; we do.
What does “read-your-writes for agents” mean? +
When one agent writes, the API returns a sequence number. Any later pack requested with that min_seq is guaranteed to reflect it — extracted, or included as a raw tail until extraction completes.
Is Lore really open source? What’s paid? +
The full server is Apache-2.0: write/read pipeline, scopes, MCP, SDKs, basic inspector. A hosted cloud and advanced governance (advanced ACL, curation, analytics) fund the project. The boundary is public — no surprises.
Can I self-host Lore? +
docker run and you’re up: one Go binary, Postgres (pgvector + pg_search), Valkey. No graph database required, no external services.
How does Lore reduce token costs? +
Packs are budget-fit and deterministic, so prompt caching compounds. A built-in meter reports tokens and dollars saved versus raw history.
How does access control between agents work? +
Policies compile to SQL predicates — enforced in the query plan, with row-level security behind it. New agents start quarantined until trusted.
When can I use it? +
We’re building in the open — the repo is live and v0.1 is the current target. Join the waitlist for early access and design-partner slots.