Policy Gate for AI Agents. Ship with approvals + cryptographic receipts.

A simple API that turns risky agent actions into an auditable workflow: propose → approve/deny → verify. Built for production teams that need control, compliance, and a tamper-evident trail.

Open the Docs Try the Live Demo

LIVE Propose actions, approve with a signed receipt, verify authenticity in one call.

Why Policy Gate?

👥

Human-in-the-loop approvals

Gate high-severity actions before anything executes. Your agents propose, humans (or policies) decide.

🔒

Verifiable receipts

Every decision produces a signed receipt you can verify later. Cryptographic proof of what was approved.

📋

Audit-ready by default

Clear trail of what was proposed, who approved, and when. Compliance teams love this.

Drop-in API

Integrate in minutes with simple HTTP endpoints. Works with any agent framework.

How Policy Gate Works

1

Propose

Your agent submits an intended action with context and severity level.

2

Decide

A human (or policy engine) approves or denies based on severity and risk.

3

Verify

Receive a signed receipt and cryptographically verify authenticity anytime.

# 1. Propose an action
POST /v1/actions/propose
→ {action_id: "abc123", status: "pending", severity: "high"}

# 2. Approve (or deny)
POST /v1/actions/abc123/approve
→ {decision: "APPROVED", receipt: {signature: "04dcfa..."}}

# 3. Verify the receipt
POST /v1/verify
→ {valid: true, message: "Signature valid. Receipt is authentic."}

Built for Real Workflows

💰 Approve outbound payments & refunds
💾 Gate database writes & schema migrations
📧 Review customer outreach before sending
Approve infrastructure changes (deploy/scale/delete)

FAQ

Is this an agent framework?

No—Policy Gate is a control layer that works with any framework. LangChain, CrewAI, custom agents—all compatible.

Can approvals be automated?

Yes—human approvals, policy rules, or both. Automation features on roadmap.

What's the fastest way to try it?

Hit /docs or run the 3-call demo above. You'll have a verified receipt in 30 seconds.

Put your agents on rails—without slowing your team down.

Start with the live API, then add gating to the actions you can't afford to get wrong.

Get Access Read the Docs