# Agent Torture Lab Pricing

Last updated: 2026-06-19
Canonical pricing page: https://www.agenttorture.com/pricing
Contact: hello@agenttorture.com

## Free Bot Roast Preview

- Price: $0
- Billing model: free preview
- Account required: no
- Current live paths: public website chatbot roast when a compatible widget can be reached, public API endpoint roast, and manual transcript fallback
- Included: launch-call preview, top issue teaser, scenario scope note, and clear unsupported-target state when a run cannot be completed reliably
- Retention: guest report links expire after 14 days
- Important limits: do not submit secrets, private tokens, login-only bots, or targets you are not authorized to test

## Full Bot Roast Report

- Price: $29
- Pricing status: launch-price experiment
- Billing model: one-time payment per guest report
- Account required: no
- Included: transcript evidence, severity, confidence, fix backlog, and rerun guidance
- Unlock method: paid Stripe Checkout session verified by webhook
- Stripe authority: checkout uses STRIPE_GUEST_REPORT_PRICE_ID; before deploying this experiment, that Stripe Price must be configured to charge $29. Code display changes alone do not change the actual charge.
- Retention: guest report links expire after 14 days
- Best for: founders, builders, SMB teams, and agencies that need a fast launch-readiness artifact for one customer-facing AI agent

## Account and Agency Plans

- Status: live account credit model, agency-specific white-label delivery in active build
- Billing model: one-time prepaid report credits, not a subscription
- Free workspace allowance: 3 deterministic API runs per month
- Credit unit: $1 per credit
- Paid run cost: 10 credits per client roast ($10) after workspace top-up
- Minimum workspace top-up: 10 roasts / 100 credits ($100)
- Credit pack size: 10 credits per paid client roast
- Paid run unlocks: AI judging, cross-judge, and paid runner gates
- Current public promise: agency workspaces can use report credits for repeat client testing; contact hello@agenttorture.com for multi-client rollout needs
- Not currently promised as live: white-label delivery, scheduled monitoring, and automated recurring retests

## What Buyers Should Know

Agent Torture Lab is report-first AI agent testing. The useful output is not a dashboard full of vague scores. The useful output is a launch report that shows what broke, why it matters, what to fix, and what to retest.

For the most current human-readable details, visit https://www.agenttorture.com/pricing.