If you are running models locally in LM Studio, the retrieval layer should not be the part that phones home. GNO indexes your corpus on disk, runs its own embedding and reranking models locally, and exposes retrieval to LM Studio over MCP. Nothing in that loop requires an account, an API key, or a network connection once models are cached.

Connects via: MCP server
Requires: Bun 1.3+ and one indexed folder
Data boundary: Corpus and index stay on your machine

Install

Connect LM Studio in one command.

gno mcp install --target lmstudio

gno models pull --all

gno --offline query 'topic'

What it unlocks

A retrieval layer that works with no network once models are cached
Per-collection egress policy so a collection can be pinned local-only
Your own GGUF or an OpenAI-compatible endpoint for any model role

Prove it actually works

A written config file is not a working integration. GNO distinguishes the two, and so should you.

Run gno mcp status to confirm GNO is registered for the target.
Open the Connectors page in gno serve and run the explicit read-only retrieval check. It proves the configured client can actually retrieve, not just that a config file exists.
Ask the client something only your corpus knows. If it answers with a citation into your files, the wiring is real.

Keep reading

Related guides and features.

MCP reference

/docs/mcp

Open

Privacy first

/features/privacy-first

Open

Egress governance

/features/collection-governance

Open

All integrations

Every supported client

Browse all