← All features

Privacy First

Corpus, index, and default models stay on disk with zero telemetry

GNO keeps its corpus, index, default model path, and opt-in retrieval receipts on your machine with zero telemetry. It works offline once models are cached. Configured HTTP inference receives role-specific text, and gno.sh receives only artifacts you explicitly publish.

Use cases
Sensitive client work and internal research
Private vaults and internal docs
Teams that need local-first AI workflows

What it gives you

  • No cloud dependency for the built-in local path
  • No telemetry or tracking
  • Fail-closed local, LAN, and remote policy per collection
  • Derived evidence retains the most restrictive source policy
  • Retrieval recording is off by default with explicit metadata or replay consent
  • Only explicit relevance labels become qrels; missing clicks never do
  • Replay recommends but never auto-personalizes ranking, prompts, models, or files
  • Local retention, export, deletion, and purge stay under your control
  • SQLite database on your disk
  • Works offline once models are cached
  • Collection root validation rejects dangerous paths
  • Concurrent writes from the CLI and MCP wait their turn instead of failing
  • Remote REST callers and every Streamable HTTP MCP caller get gno:// URIs and collection-relative paths, never host file paths
  • CSRF protection on API endpoints

Try it yourself

Representative commands and entry points. Full reference lives in the documentation.

gno doctor
gno status
gno collection policy get notes
gno trace list --md

Keep reading

Related features and docs.