How it works

From scattered to searchable.

Deployment is fast, low-friction, and doesn't require your engineers. Three steps take you from scattered documents to a live, cited intelligence layer.

01

Connect your data sources

Point Clag at the systems your knowledge lives in — Notion, Confluence, Google Drive, Slack, databases, email, folders of PDFs. It ingests and indexes everything across your stack, and keeps the index in sync as your content changes. Existing permissions are preserved, so people only ever retrieve what they're already allowed to see.

02

The RAG pipeline is deployed

Clag builds vector embeddings of your content and stands up a retrieval layer on top of the language model, grounded entirely in your context. Every question first retrieves the most relevant passages from your own data, then the model reasons over those — not the open internet — to compose an answer.

03

Your team goes live

People query through chat, the REST API, or the tools they already use. Every answer comes back with citations to the source documents, so it's verifiable. We monitor quality and tune retrieval as your knowledge base — and your team — grows.

Where it runs

On your infrastructure, not ours.

Clag deploys via Docker Compose on your own server, next to a local vector database. All credentials, employee data, and documents stay local — the only outbound call is a license check. Air-gapped deployments work end to end.

See what it can do.

Private deployment, cited answers, enterprise access controls, and more.

Explore capabilities →