Give an agent durable memory
Write typed facts as an agent works, then retrieve them with natural-language hybrid search or an MCP tool belt. See the agent-memory guide.
little big brain
An object-storage-native temporal graph and hybrid-search engine for AI context — lexical, vector, ontology, graph paths, facets, and hybrid ranking over one snapshot.
What you can build
Section titled “What you can build”little big brain stores graph facts as append-only events, builds search indexes as disposable object-backed accelerators, and runs semantic hybrid search across BM25, vector, ontology, and graph-neighborhood signals over the same snapshot. It is a full search engine — not just a graph store — with an append-only temporal model, provenance, SPARQL/SHACL query surfaces, and a hosted SaaS you talk to with a stack API key.
Hybrid retrieval over your knowledge
Fuse BM25, vectors, ontology expansion, and graph neighborhoods with filters and facets. See hybrid retrieval.
Query & reason over a temporal graph
SPARQL 1.1, a structured SELECT/aggregate surface, and SHACL shapes over an append-only, snapshot-consistent graph — plus reasoning that derives facts you never wrote. See SPARQL & SHACL.
Tune retrieval with real feedback
Grade results, export qrels, and fine-tune managed embeddings. See embeddings & feedback tuning.
Pick your entry point
Section titled “Pick your entry point”TypeScript — @lbb/clientDependency-free fetch wrapper for Node, browsers, and edge workers.
Python — lbbSync + async httpx client with typed Pydantic models.
MCP server — @lbb/mcpA graph-aware tool belt for Claude, Cursor, and Codex — local or hosted.
HTTP API & SPARQLThe raw REST surface plus the native SPARQL 1.1 Protocol endpoint.
ConsoleThe web app at cloud.littlebigbrain.com: explore, query, index, and operate.
New here?
Section titled “New here?”Core conceptsSnapshots, WAL, hybrid search, consistency.
Authentication & endpointsStacks, API keys, and base URLs.