AI cover is priced once a year from a questionnaire nobody can check. The Insurer Feed gives an insurer each policyholder's live Insurability Level, worked out from what their AI actually did and sealed into Bitcoin. A signed alert arrives the moment it changes, a sealed statement arrives every month, and claim evidence takes one click instead of a six-month investigation.
Every decision in the last 30 days sealed, every record verifies, no missing receipts.
Humans review the flags and take real time to read them. No sub-two-second rubber stamps, no reviewer who never disagrees.
An armed enforcement loop and a kill switch, so a rogue AI is cut off the moment it crosses the line.
AI agents can't spend money without a sealed approval within set limits.
Levels are recalculated from behaviour every hour. A drop only counts after 3 days in a row at the lower level, so one bad deploy is a warning with evidence, not a penalty. Altered records count at once.
Behaviour-based pricing, the way black-box car insurance works. You set the price and write it into your policy. sebbi.pro supplies the evidence you can price from.
A signed webhook fires when a policyholder's level changes, with the evidence and a sealed block reference.
Missing receipts, altered records, fast approvals, every kill-switch trip and release with timings. One sealed report for the claim period.
Ask whether a policyholder met your required level at the exact moment of a loss. You get MET, NOT_MET or UNKNOWN, sealed, with the facts behind it. It pairs with our model policy wording.
You only see policyholders who share with your domain. They can revoke, and every share and revoke is sealed.
Everything here has an API. The full spec is at https://sebbi.pro/.well-known/insurer-feed.json. AI agents can use the MCP tools sebbi_insurer_level, sebbi_insurer_share and sebbi_claim_evidence at https://sebbi.pro/mcp.
sebbi.pro supplies evidence. Insurers set price and decide claims under their own policy terms.