Documentation
Zealoop is a support agent that answers from your own documentation, reads the signed-in customer's own records, and takes actions you define — with the guards enforced in code rather than asked for in a prompt.
Quickstart
From an empty workspace to a grounded answer on your own site, in about twenty minutes.
The turn pipeline
Six stages, always in the same order. What each one costs and when it refuses.
Knowledge
Crawl a site, upload a file, paste a snippet. How chunking and retrieval actually work.
Identity & signing
Why an unsigned email is a claim, not a fact — and how to sign one properly.
Tables
Give the agent customer rows it can look up, keyed on an identity column.
REST API
Every dashboard endpoint, the auth model behind it, and the widget surface.
What it is
You give Zealoop three things: documentation to answer from, optionally some customer data to look up, and optionally some API endpoints it may call. It gives your visitors a chat widget that resolves their problem instead of listing five articles that might contain the answer.
Every workspace is isolated. Knowledge, tables, actions, conversations and traces are scoped to an organization, and no query crosses that boundary.
What makes it different
- It abstains. When reranking puts every candidate below the confidence threshold, the agent says it doesn’t know instead of assembling a plausible paragraph. That is a configured threshold, not a hope.
- It shows its work. Every turn writes a trace: the rewritten query, the candidates and their scores, which model ran, token counts, cost, latency per stage, and the outcome. You can read exactly why any answer happened.
- Its guards are code. Identity, confirmation and test-pass requirements are checked in the execution path, not described to the model. A prompt injection that convinces the model still gets refused by the code that runs the call.
Where to start
If you have twenty minutes, follow the Quickstart — it takes you from an empty workspace to a grounded answer on your own site. If you are evaluating whether this fits your architecture, read The turn pipeline first, then Actions.