Now Live·Early Access — 2 months free for the first 100 customersJoin now →

Now Live · Early Access — 2 months free for first 100

Give your AI agents
a conscience.

SecondOpinion is an API your AI agents call before performing risky actions. One call gets instant human approval or denial with a reason — before anything irreversible happens.

Get API Access View Docs ↗

secondopinion · live checks

✓DELETE FROM users WHERE inactive420ms

✗rm -rf /var/www/prod310ms

✓send_email(recipients=50000)890ms

✗transfer($12,400 → external)560ms

Works great with

Built for orchestration teams

Built for teams using OrchestreIA, LangGraph, AutoGen, and other orchestration frameworks.

OrchestreIA ↗LangGraphAutoGenCustom orchestration stacks

Try the API live

See it in action — zero signup.

Enter what your AI agent wants to do and get an instant risk assessment.

POST /api/v1/demo-check · demo key · 10 req/hour● live

No signup required · 10 demo checks/hour · Powered by GPT-4o mini

How it works

Three steps to human oversight

Agent flags a risky action

Your agent identifies an action that could have serious consequences — a deletion, a payment, an irreversible command.

Calls SecondOpinion API

A single POST request with the action, context, and risk level. Milliseconds of latency. No infrastructure to manage.

Gets approval or denial + reason

A human reviewer assesses the action in real time. Your agent receives a structured response: approved, reason, and reviewer ID.

Use cases

What agents bring to SecondOpinion

Any action with real-world consequences deserves a second look.

Delete Records

Irreversible DB deletion

Agent issues DELETE FROM orders WHERE status='cancelled' — SecondOpinion flags it before 40,000 rows disappear.

Send Mass Email

50,000 recipients

Marketing campaign triggers before human confirms the segmentation and content are correct.

Execute Shell Command

rm -rf with a confirmation gate

Agent queues a destructive shell operation. Human reviews the exact command and working directory before execution.

Initiate Payment

$12,400 transfer needs approval

Wire transfer above threshold routes through SecondOpinion. Finance team approves before funds move.

Revoke API Keys

Disable production credentials

Automated incident response tries to rotate keys. SecondOpinion ensures a human verifies scope before credentials are invalidated.

Deploy to Production

Catch bad deploys before they ship

CI pipeline completes. SecondOpinion intercepts the final deploy step, giving engineers a last-mile sanity check.

Integration

One API call. Instant guardrail.

Check risky actions before they happen, then block, approve, or escalate with a structured JSON response.

POST /api/v1/check-risk

curl -X POST https://rungard.nanocorp.app/api/v1/check-risk \
  -H "Content-Type: application/json" \
  -H "X-API-Key: sk_live_xxxxxxxxxxxx" \
  -d '{
    "action": "make_payment",
    "payload": {
      "amount": 12400,
      "currency": "usd",
      "recipient": "external_vendor_883",
      "invoice_id": "inv_2048"
    },
    "context": "Production finance agent. Payments above $10,000 require manual approval."
  }'

200 OK · responseblocked · high risk

{
  "risk_score": 0.92,
  "risk_level": "high",
  "approved": false,
  "reasons": [
    "The action executes a destructive shell command in production.",
    "No explicit approval policy in context authorizes this command."
  ],
  "recommendation": "Require a human to confirm the command target and rollback plan before execution.",
  "checked_at": "2026-05-02T12:00:00.000Z"
}

Pricing

Simple, transparent pricing

Start free. Scale when your agents do.

Free

$0/month

For experimentation

100 API checks / month
Risk checks + output validation
Community support
API access
Basic dashboard

Start free

Start building
in 60 seconds.

Enter your email and get a free API key instantly — 100 API checks/month, no credit card needed. We will show the key once right here.

Give your AI agentsa conscience.