Block risky agent actions before they execute

Human-in-the-loop enforcement at the primitive boundary.

Read the launch post →
audit log$250 cap
  • 10:14cards.createAllowed
  • 10:22payment.chargeAllowed
  • 10:23payment.chargeRevoked

Paste and ship

Drop this into Cursor, Claude Code, or any agent. It reads the skill manifest and wires up Approvals for you.

Agent prompt

Read https://usenaive.ai/skill.md and use the Naïve Approvals primitive in my project. Start with: naive approvals list

What you get

Enforce before executionAgents wait — they don't bypass policy.
Policy rules in codeThresholds and actions in agent.config.ts.
Approval queueSlack, email, or your own UI.
Audit trailWho approved, who denied, full replay.

One call to approvals

CLI, SDK, or REST — one bearer token, one credit balance.

CLI

Approvals

High-risk primitive calls never reach the wire until a human approves — large spends, formation submits, trading orders. Denied actions are blocked, not logged after the fact.

Enforce before execution
Agents wait — they don't bypass policy.
Policy rules in code
Thresholds and actions in agent.config.ts.
Approval queue
Slack, email, or your own UI.
Audit trail
Who approved, who denied, full replay.
# Run with your API key exported
$ naive approvals list

Start building with Approvals