nanograph

On-device graph database for agents and humans. One CLI. One folder. Schema-as-code. No server.

Built on Rust, Lance, Arrow, and DataFusion.

Website | Docs | Examples

Why nanograph

• On-device -- no server, no cloud, no Docker. Everything stays on your machine.
• Schema-as-code -- .pg files version-controlled in git, enforced at query time. No "property not found" at runtime.
• Built for agents -- Claude reads, writes, and traverses the graph natively.
• Fast -- Rust + Lance + Arrow columnar execution. Sub-millisecond opens, ACID, time-travel.
• Full search stack -- full-text, semantic, fuzzy, BM25, hybrid, graph-constrained reranking.
• Built-in CDC -- every mutation logged to a ledger. Replay, audit, sync from any version.
• Zero setup -- create a graph and start querying. Delete and recreate in seconds.

Install

brew install nanograph/tap/nanograph

Or from source (requires Rust 1.85+ and protoc):

cargo install nanograph-cli

SDKs

• TypeScript/Node.js -- npm install nanograph-db
• Swift -- Swift Package via the nanograph-ffi wrapper

Quick start

nanograph init my.nano --schema schema.pg
nanograph load my.nano --data data.jsonl --mode overwrite
nanograph check --db my.nano --query queries.gq
nanograph run --db my.nano --query queries.gq --name my_query

See examples/starwars/ for a ready-to-run demo with 66 nodes, 146 edges, and 30 queries.

Use cases

• Context graphs and decision traces for AI agents
• Agentic memory with typed, sub-100ms local queries
• Personal knowledge graphs with schema enforcement
• Dependency and lineage modeling
• Feature generation for ML pipelines

License

MIT