Pick the documents each agent uses.
Agents use only the documents your administrators add. Choose which collections each agent can search, and add key facts it can answer from without searching.
Every answer shows its sources, and administrators can read the instructions each agent followed.
Administrators can reviewConversationsInstructions followedResponse times and errorsAI product access log
Agents use only the documents your administrators add. Choose which collections each agent can search, and add key facts it can answer from without searching.

Set each agent’s rules with simple options: topics to avoid, what to do when your documents don’t have the answer, and whether it may search the web. Then pick its model and tools.

Every answer lists its sources, so anyone can check the passage. Administrators can read past conversations and the instructions the agent followed.

Only people you add can sign in. Choose which people and teams can use each agent.
Run a model on your own servers, or use one from OpenAI, Anthropic, or Google. Each agent can use a different one.
Claude, ChatGPT, and other AI products connect over MCP, and each person signs in with their own account through OAuth 2.1. Same permissions, every use logged, and you can disconnect at any time.
Groundbase and every AI model it uses can run on your own servers. Nothing leaves your network unless you connect an outside service.
AI running on your serversAnswersDocument searchResult ranking
Your documents are always indexed on your own servers, whatever outside services you connect.
Every outside service is used only with your permission.
Tell us what your team needs to find and do, or what you’d like to offer your customers. We’ll reply by email.