Your next customer might be an agent. Nobody tests for them.
alt.qa exists because the fastest-growing visitor on the web, an AI agent acting for a person, is the one no release process checks. We run those agents, and the people they act for, through your funnels before you ship.
Why we built it
Readiness scanners check files that agents mostly ignore. Monitoring tools tell you after the sale is lost. Your QA suite tests one idealised user of your own app. None of them runs the real agents, from every vendor, through your real funnel and shows you where they stall.
We started with browser tools for capturing bugs (Scan) and load (Hit). In September 2026 we narrowed everything to one job: Agent Readiness Assurance.
What we will not do
- Impersonate ChatGPT Agent, Comet, Muse or any consumer agent. We use real vendor models under our own signed identity and call the classes what they are: behavioural approximations.
- Say you are “compliant”, “certified” or “agent-ready”. We publish evidence, not badges.
- Claim our personas are older, disabled or non-native people. They are simulations with published precision, confirmed by paid testers from those communities.
- Complete a real payment. Every run stops before pay unless it is in a sandbox.
- Consent first: only properties you own or authorise
- Honest identity: signed, never spoofed
- Stop before pay, always
- Evidence, not empathy
- Precision published with n and confidence intervals
alt.qa is a standalone company. eval.qa supplies paid human verifiers (the Eval Army) under an arm’s-length services agreement. werify.ai signs our evidence (hash, timestamp, chain of custody). re.gress.ai is a separate CI guard for your own prompts and models.
Tell us about your funnel.
We will tell you honestly whether an Alt Matrix would find anything worth fixing.