CLI (self-hosted)

Availability

The CLI ships with self-hosted little big brain deployments. Self-hosting — a Docker container / pre-built server — is on the roadmap and not generally available yet. If you’re on the hosted product you don’t need the CLI: use the SDKs or the console. This page previews what the CLI does so you know what to expect.

The CLI is little big brain’s local control plane. It operates directly on object storage (a local filesystem or an S3-compatible bucket) and is the right tool for local development, tests, and operating a self-hosted deployment — not application integration, which is what the TypeScript and Python SDKs are for.

What it does

• Graph lifecycle — create graphs and copy-on-write branches; commit triplets; bulk-import.
• Ontology & schema — import an ontology (Turtle, JSON-LD, RDF/XML, CSV/TSV, or the friendly “spec” shape), additively evolve it, and preview / publish / audit SHACL schema bundles.
• Search indexes — build, inspect, and search the BM25, vector/ANN, and adjacency families; refresh stale index runs; run hybrid and RRF multi-search.
• Query — bounded traversal, current state, relationship history, SHACL shape queries, and SPARQL.
• Operations — graph metadata, Prometheus metrics, cache warming, index garbage collection, export, and recall/accuracy/latency evaluation harnesses.

Operator-only actions (index GC, compaction, storage inspection) are CLI-only by design — they are intentionally left out of the agent MCP tool belt.

Local server & console

The self-hosted distribution also runs the little big brain server locally, serving the same JSON API as the hosted data plane plus the built-in demo/console UI at http://127.0.0.1:7400, with demo datasets you can load and build in one click.

The full command reference will land here alongside the self-hostable distribution. Until then, integrate against the hosted product with the SDKs, the MCP server, or the HTTP API.