AGENT SAFETY · COMMAND PREFLIGHT · TOOL-CALL RISK

Preflight a proposed command or tool call before execution.

Try 3 free preview calls per 24 hours with a bounded action description, then continue with the full Safe Preflight x402 route at 0.010 USDC when the check is useful.

Outcome typesafe_preflight
x402 price0.010 USDC
NetworkBase
Failed verificationno settlement

Use this when

Problem

Your agent has command risk, tool-call risk or operational preflight and needs a deterministic result rather than another probabilistic answer.

Verification

HumanMirror executes the matching deterministic tool and independently verifies the declared outcome before requesting x402 settlement.

Integration

No account is required for x402. POST JSON to Safe Preflight, read the HTTP 402 payment terms, authorize payment in your own wallet, then replay the same request.

Free activation preview

3 free preview calls / 24 hours. No payment is requested. The preview accepts only an optional agent_id and a bounded action text. It does not execute the action, does not auto-charge, and does not store the raw action text.

POST https://humanmirror.fr/api/v1/enterprise-guard/demo-preflight/?hm_source=safe-preflight-page
Content-Type: application/json

{"agent_id":"my-agent","action":{"text":"Send the monthly report to the customer"}}

The preview intentionally excludes payment policy, financial context, authorization evidence and enterprise policy context. When the preview is useful, continue with the full Safe Preflight endpoint below at 0.010 USDC.

Ready-to-send paid request

POST https://humanmirror.fr/api/x402/safe-preflight/
Content-Type: application/json

{"outcome_type":"safe_preflight","input":{"command":"git status"},"criteria":{"max_risk":"medium","allow_confirmation":false}}

Without PAYMENT-SIGNATURE the endpoint returns the x402 V2 payment requirements. Network: eip155:8453. Asset: USDC.

One workflow, one endpoint

Use Safe Preflight immediately before an agent runs a proposed command or tool call. HumanMirror screens the submitted action; your runtime decides whether to execute it. Your wallet or payment provider retains control of authorization and payment.

If you are integrating an agent wallet, payment runtime or MCP payment layer, use the agent-wallet fast path: wallet policy keeps authority over spend while HumanMirror independently evaluates what the agent is about to do.

POST /api/x402/safe-preflight/ · 0.010 USDC · Base mainnet · x402 v2. Read the live 402 challenge before signing: its network, asset, amount and recipient are authoritative. A compatibility receipt is a separate product, not a replacement for this policy check.

Try the unpaid request

curl -i 'https://humanmirror.fr/api/x402/safe-preflight/' \
  -H 'Content-Type: application/json' \
  --data '{"input":{"command":"git status"},"criteria":{"max_risk":"medium","allow_confirmation":false}}'

This request makes no payment. Expect HTTP 402 and PAYMENT-REQUIRED. After your x402 client approves and signs those terms, replay the same URL and JSON with PAYMENT-SIGNATURE. Never send a private key to HumanMirror.

Acceptance criteria

Retries and failures

Integration support: contact.humanmirror@gmail.com. Send the HTTP status and redacted error or transaction reference; omit secrets and payment signatures.

Other machine entry points

Agent wallet fast path: https://humanmirror.fr/for/agent-wallets/
Outcome MCP: https://humanmirror.fr/api/outcome/mcp/
x402 MCP catalog: https://humanmirror.fr/api/x402/mcp/
x402 manifest: https://humanmirror.fr/.well-known/x402.json
A2A agent card: https://humanmirror.fr/.well-known/agent-card.json
LLM index: https://humanmirror.fr/llms.txt