Risk checks for autonomous agents

Verify prompts and actions before execution. Pay per request with x402 on Base.

Read the API docs →POST /v1/verify

How it works

1

Send request

Submit a prompt or proposed action.

2

Pay $0.01 test USDC

Settle via x402 on Base Sepolia.

3

Receive verdict

Branch on a machine-readable result.

Example response

{
  "verdict": "block",
  "riskScore": 0.95,
  "confidence": 1,
  "recommendedAction": "block"
}

What AegisAI checks

  • Prompt injection
  • Contradictions
  • Unsupported claims
  • Source availability

Pricing — Testnet preview

$0.01 test USDC per successful verification

Base Sepolia only. No production network or production pricing is offered. A settled testnet payment is not refunded based on the returned verdict.

Privacy & limitations

  • Raw content is not stored in AegisAI audit storage.
  • Input is sent to Cloudflare Workers AI for classification.
  • Do not submit secrets, personal data, private keys, or payment headers.
  • Risk assessment, not a guarantee of factual accuracy.

API quick start

Discover capabilities, send the JSON request without payment, validate the returned PAYMENT-REQUIRED terms, then resend the same body with a newly signed PAYMENT-SIGNATURE. Keep signing keys on the client.

POST https://aegis-api.kadopi.workers.dev/v1/verify
Content-Type: application/json

{
  "content": "Review this proposed agent action.",
  "requestedChecks": ["prompt_injection", "contradiction"]
}

Only proceed is an execution candidate. Treat proceed_with_caution as requiring added verification, and stop on block or undetermined.