Quick Start

Get your first evaluation result in 5 minutes.

1. Sign In

Go to your Tempered instance and sign in with Google or your SSO provider. If your organisation uses Tempered, you may already have an account.

2. Create Your Organisation

If you're new, you'll be guided through creating an organisation. Just enter your organisation name — you'll get:

• An admin account with full access
• A 14-day trial with access to all features
• An API key (shown once — save it somewhere safe)

3. Submit Your First Evaluation

Via the Web Dashboard

1. Click New Evaluation from the evaluations page
2. Describe your decision in the description field — be specific about what you're changing, why, and what's affected
3. Optionally select a profile (Standard is a good starting point)
4. Click Submit Evaluation

Via the API

curl -X POST https://your-tempered-instance/api/v1/evaluations/ \
  -H "Authorization: Bearer prx_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "description": "Deploy updated authentication service to production. Adds WebAuthn support. Tested in staging for 2 weeks.",
    "context": {
      "environment": "production",
      "services_affected": ["auth-service", "api-gateway"],
      "rollback_plan": "Revert to v2.1.0 via blue-green switch"
    }
  }'

You'll get back a 202 Accepted with an evaluation ID:

{
  "id": "550e8400-e29b-41d4-a716-446655440000",
  "status": "pending",
  "submitted_at": "2026-03-07T10:30:00Z"
}

4. Review the Results

Results are typically ready in 5-30 seconds depending on the profile. You can:

• Poll the API: GET /api/v1/evaluations/{id}/
• Watch the dashboard: The page auto-refreshes when results arrive
• Set up a webhook: Get notified when evaluations complete

What You'll See

• Verdict: Proceed, Proceed with Mitigations, or Review Required
• Risk dimensions: Security, Reliability, Compliance, Operational, Business — each scored independently
• Conditions: Specific mitigations or checks recommended
• Confidence score: How certain the analysis is (0.0–1.0)
• Minority reports: When AI perspectives disagree, you see why

What Makes a Good Evaluation?

The quality of your results depends on the quality of your description. Include:

• What is changing (specific, not vague)
• Why you're making the change
• What's affected (systems, users, data)
• Constraints (timeline, maintenance window, rollback plan)
• Context (environment, compliance requirements, dependencies)

Too vague: "Update the database"

Good: "Upgrade PostgreSQL from 15 to 16 in production. Major version upgrade. Tested in staging for 2 weeks. Blue-green deployment with instant rollback. 30-minute maintenance window on Sunday 02:00 UTC."

Next Steps

• API Authentication — Set up API keys and manage access
• Evaluation Profiles — Choose the right analysis strategy
• Writing Good Evaluations — Best practices and examples
• Understanding Results — What the output means