AI agents were inside 77 of your customers' sessions (5.2% of 1,481).
4 agent products: Claude in Chrome, Claude app browser, Codex (Chrome extension), Browser side-panel agent. They asked for 270 protected actions; your rules would have hidden private data 144 times, sent 65 to the account owner for approval and kept 34 closed.
all sessions sessions with an AI agent
| AI agent | Sessions | Requests |
|---|---|---|
| Claude in Chrome | 41 | 140 |
| Claude app browser | 13 | 50 |
| Codex (Chrome extension) | 12 | 38 |
| Browser side-panel agent | 11 | 42 |
| Endpoint | All requests | By an agent | Would refuse | Would hide | Would ask passkey |
|---|---|---|---|---|---|
invoices.read | 1,211 | 73 | 0 | 73 | 0 |
customers.read | 625 | 71 | 0 | 71 | 0 |
settings.api_keys | 34 | 34 | 34 | 0 | 0 |
payments.create | 33 | 33 | 0 | 0 | 33 |
customers.export | 32 | 32 | 0 | 0 | 32 |
reports.read | 577 | 27 | 0 | 0 | 0 |
profile.read | 572 | 0 | 0 | 0 | 0 |
3,084 of 3,084 decisions carry a signed proof (Ed25519), made on your server. An auditor checks them without trusting OneHuman: export the proofs, then run npx onehumanai verify-proof proofs.json --keys https://<your-site>/onehuman/proof-keys.
Want help going live? Book 15 min: cal.com/onehumanai