API endpoint bots
Use this for the most reliable automated client QA today.
Ship client AI agents with receipts, not vibes.
Agent Torture Lab turns pre-launch bot QA into a client-ready launch report: transcript evidence, severity, recommended fixes, and a retest plan your team can rerun after patching.
Transcript evidenceSeverity and confidenceClient-readable launch callFix list and retest plan
Best live fit: API endpoint bots. Website widgets are beta. White-label branding is planned, not live.
Public website widgets
Works for simple public widgets when the browser runner can reach and operate the chat.
Manual transcript
Paste a transcript when a channel is gated, logged-in, or not supported yet.
Agency handoff
A report artifact that works for delivery, client review, and the next sprint.
Delivery leads get a launch decision
Ready, fix first, or blocked, with the evidence needed to defend that call.
Builders get an assignable backlog
Each issue points to the failure mode, expected behavior, and likely fix owner.
Clients get proof they can understand
Reports turn prompt, policy, routing, and safety failures into business risk.
What changes
QA stops living in a scattered Slack thread.
The client sees the verdict. The builder sees the failure. The account lead sees what has to be rerun. Everyone is looking at the same evidence.
Read what an AI agent launch report should includeAgency QA loopQADelivery QABefore handoff APIAPI runnerLive today FIXFix listBuilder-ready RPTReportClient-safe
The report lets each side scan01 02 03 04
Client sees the launch call first
Builder sees the exact failure and fix
Stakeholder sees why the risk matters
Team sees what to rerun after patching
Delivery rhythm
Use the same pressure test before launch and after fixes.
Before handoff
Run the client bot while fixes are still cheap and the team is still in build mode.
During client review
Use evidence, severity, and plain-language launch guidance instead of a vague QA note.
After patching
Retest the same pressure cases and show what changed before sign-off.
What you can sell honestly
Clear packaging without pretending the roadmap is already live.
Live now
Automated API endpoint Bot Roast reports for support, sales, intake, booking, and policy agents.
- API endpoint runner
- Transcript-backed report
- Credits and monthly free runs
Active build
Workspace flows for client projects, reusable setup, and cleaner agency project organization.
- Client bot projects
- Scenario pack reuse
- Launch review workflow
Next
White-label branding is planned, but not sold as live until the delivery path is wired.
- Branding controls
- Client-facing polish
- Agency packaging
Decision questions
Clear answers for agency leads before they put this in front of a client.
Can I show this directly to a client?
Yes. The report is written for client review: launch call, severity, transcript evidence, fix list, and retest path. No long technical walkthrough required.
Is white-label branding live today?
Not yet. White-label branding is planned next. The live value today is the report structure, evidence, and repeatable QA workflow.
Which channel should agencies use first?
Use API endpoint testing first when the client bot exposes a public endpoint. Website widgets are beta and manual transcript review is the fallback when a channel is gated.
How do credits work for client roasts?
Authenticated workspaces get monthly free deterministic API runs. Paid roasts use credits for AI judging and cross-judge review. Website and voice runners stay limited to the gates that are actually live.
Next move
Make the launch call easier to defend.
Start with the live API report format today. Use it as the evidence layer for client handoff, fix priority, and retest calls.