EntityAgent is a public answer-agent runtime grounded in a project's approved, published snapshot. Requests require an enabled project, active entitlement, signed allowed origin, bounded input, rate and cost capacity, and a configured provider. The system prompt and retrieval context restrict answers to approved evidence and should fail safely when support is missing, but no generative model is described as incapable of error; responses still require monitoring and correction paths. The runtime and embed boundary are repository-implemented, while the separately priced self-service add-on remains contained until its purchase, renewal, failure, refund/dispute, and reconciliation lifecycle is complete.
EntityAgent is constrained to approved content, unlike a general chatbot
EntityAgent differs from a general chatbot by restricting retrieval and instructions to the approved published snapshot. That materially narrows risk; it is not an absolute accuracy guarantee.
EntityAgent deploys with a single script tag on any page
The intended embed uses a script/widget tied to a signed origin, project, entitlement, and published snapshot. Do not enable it until the production signing secret, origin matrix, provider, and runtime evidence pass. Learn more on the EntityAgent page.