AI inside your product. Your brand, your invoice, our infrastructure.
For product leaders at SaaS companies, vertical ERPs and booking platforms: the agent layer without building an AI team. This page is the technical contract — with real dates, not demos.
The truth first
None of the six pieces exists today as a public API — which is why each carries a date, just like the guarantee table at /enterprise. What does exist today: the multi-tenant platform that will serve them (RLS isolation, approval-gated versioning, evals), and a first design partner integrating with us. If your timeline matches ours, you get a voice in the API design.
Signed outbound webhooks
HMAC, retries with backoff and a published event catalog. Your product knows everything the agent does.
OpenAPI spec of the connector
The connector contract with your product — 70% of the integration work, specified before writing code.
Conversation read API
Render the agent’s conversations inside your own UI, in your own design.
Human-handoff event and endpoint
Escalation to a person lands inside your product — your inbox, your agents, your flow.
Per-customer usage API
How much each of your customers consumes, for your billing and your margins.
Per-customer spend cap
A hard per-customer limit: no tenant of yours can hand you a surprise bill.
Same as /enterprise: dated items go into the contract with their date, and a no-penalty termination right if missed.
- Price curve to 500 customers
- Directional and tiered: per-customer price drops with volume. Figures open up in the engineering call — full channel economics are under pricing review and we do not publish numbers that may change.
- Sandbox isolated from production
- A test environment whose test keys never touch production — a WP-46 prerequisite, sharing its date. Until then, design integrations run hand-in-hand with engineering.
- Published versioning and deprecation
- The API ships with a written versioning policy, deprecation windows and rate limits with numbers — no product leader integrates against an API without a change contract.
Would you embed AI if you did not have to operate it?
Tell us your product and your timeline. If there is a fit, the first call is with the engineer building the API — and your case shapes its design.