QUESTIONS
How it works, in plain words.
What is STEWARD?
STEWARD is an embedded AI product expert delivering verifiable outcomes. You point it at your documentation and your API. Your users get an expert inside your software that has read everything, checks their actual state, refuses to guess, and files a record of every answer.
Is STEWARD a chatbot?
The panel on this page is a rendering surface with no logic of its own. STEWARD is the service behind it. It reads your material, operates your API for the user, refuses when the material does not support an answer, and files a record. You can call the API directly and receive the same answer with the same trace.
What does STEWARD read?
The documentation and data sources you connect, and your API when a question needs the user's actual state. It answers only from that material and it cites the source beside the answer.
What happens when it does not know?
It says so, names what it did not find, and hands the question back to the person. The refusal is designed behaviour, and it is recorded like any other answer.
Where does my data go?
Your corpus is scrubbed of personal data before indexing. Your users' credentials pass through to your API and are never stored. Inference runs on AWS Bedrock with per-product isolation. Your users' questions and live API responses do pass through the model when the question requires it, and personal-data redaction on output is a hard fail.
Which model does it use?
Inference runs on AWS Bedrock with per-product isolation. The model predicts text. STEWARD does the reasoning, the guardrails, and the record.
How do I embed it?
Sign in at steward.link, create a product, connect your documentation, mint a publishable key, and paste the snippet. STEWARD fits any container on any page: a launcher button, a panel that slides in from the side, or a fixed area in the body of the page, like the one above.
What does it cost?
Knowledge-base answers are free on every tier. The meter runs when the expert does expert work, and your invoice reads from the trace. The pricing page has the numbers.
Do you have SOC 2?
Independent security review, July 2026. OWASP SAMM Level 2 with several Level 3 characteristics. About 1,200 automated tests on every build. SOC 2 Type II is in progress.
Does STEWARD identify itself as AI?
Yes. STEWARD identifies itself as an AI system on every surface, and its outputs carry machine-readable AI marking, as Article 50 of the EU AI Act requires.
Can I try it before I sign up?
Yes. Ask STEWARD about STEWARD on this page, then open the trace of the answer you received.