Commt

FAQ

What is a private AI agent?

A private AI agent is one that reaches only the knowledge bases, tools and connectors you explicitly bind to it. You set that scope when you create the agent, choosing which knowledge bases it reads from and which tools it can call, and every request stays inside that scope.

What data does Commt store?

By default, Commt does not store message content, only metadata for your usage report: request counts, latency and which agent answered. Message text is written only when you switch on a debug window or when a conversation is flagged as a failure. You set your own retention period for what is kept, from nothing up to a year.

Which models can I use?

Commt agents can use frontier models from Anthropic, OpenAI and Google, plus open-source models such as Llama and Qwen. Your plan sets which models an agent can reach, and switching a model is a dropdown change, not a migration, so instructions and knowledge bases carry over unchanged.

Where is Commt hosted?

Commt is hosted in the EU. Application, database and document storage run in the EU on private networking, with your files in isolated object storage and your vectors in an isolated database, neither reachable from the public internet.

Can a Commt agent access the internet?

Yes, by default: a Commt agent can search the web and open pages on every plan, alongside any knowledge bases, tools and connectors you have bound to it. Switch that off when you create the agent to keep it a private AI agent that reaches only what you have explicitly scoped to it, such as Slack or Jira, each limited to the service it connects to, and nothing on the open web.