Approvalsgateway
Block risky agent actions before they execute
Human-in-the-loop at the tool boundary. Risky calls freeze until someone decides.
Allowed, or revoked. Logged either way.
audit log$250 cap
- 10:14cards.createAllowed
- 10:19email.sendAllowed
- 10:22payment.chargeAllowed
- 10:23payment.chargeRevoked
- 10:31formation.submitAllowed
- 10:38trading.orderRevoked
- 10:44vault.revealAllowed
Paste and ship
Drop this into Claude Code, Codex, or Cursor. It reads the docs and wires up Approvals for you.
Agent prompt
Read https://vetta.sh/docs/concepts/policies and use the naïve Approvals primitive in my project. It is part of the gateway layer; wire it through naive.config.ts rather than calling a vendor directly.
Used by
- AI Media ChannelA channel where every video is original: briefed, generated in the channel's own look, and queued for you by an agent team on a schedule.View →
- Clipping ChannelLong-form in, clips out. The moments worth cutting found, reframed, captioned, and posted on cadence.View →
- AI Automation AgencyThe whole agency machine with no specialism attached: CRM, public site, a workspace per client, and a pair that works the pipeline. You bring the trade.View →
- SEO / GEO AgencyRanks clients in search and in AI answers. CRM, a crew per client, and a public site out of the box.View →
- NewsletterA publication that reads its beat, drafts the issue, and sends on schedule while it grows its own list.View →
- Paid Ads AgencyCreative produced, campaigns launched, budget moved daily toward what returns. Under a cap agents can't raise.View →
- Recruiting AgencyRoles taken in, candidates sourced and screened, shortlists scored against the brief requirement by requirement.View →