Agent Usability: know the day an agent stops getting in
Save one task an agent should be able to do in your product, such as signing up and making its first API call. Tanso runs a real agent through it on a schedule, checks the result itself, and emails or webhooks you when it stops passing.
Checks are free. Hosted test runs: 5 free a month, then $5.00 per completed run. Inconclusive runs and runs that fail on our side are free. Sign in to the hosted dashboard.
Keep one task working, release after release
- Save the task: name the task, your product's address, and the response that proves it worked. Tanso reads that response after the run instead of trusting the agent's own account.
- Run it on a schedule: a real agent runs it on Tanso's model, daily or weekly, or whenever you start it. Each run is compared with the last, so a new failure shows up as a regression.
- Hear when it breaks: once the test has passed, you get an email, a signed webhook, or both when it stops passing, and again when it passes.
What it checks
Each test proves one step an agent takes. A free check reads your public pages and says which steps they describe; a test run proves a step by sending a real agent through it.
| Area | Step | Passes when the agent |
|---|---|---|
| Web | Discover | finds the product and its docs in machine-readable form |
| Web | Understand | reads what the product does and what it costs without scraping HTML |
| Onboarding | Sign up | creates an account on its own, or through a documented handoff to a person |
| Onboarding | Access | holds a scoped key that works on its own data and is refused on anyone else's |
| Monetization | Use | makes a real call with its key |
| Monetization | Pay | completes a charge or subscription without a browser |
| Monetization | Manage | changes a plan or cap and sees the change take effect |
In the hosted dashboard, each step shows one of: agent did (a test run proved it), agent can (public pages describe a way for an agent to do it alone), handoff (a person steps in at a documented point and the agent carries on, which counts as passing), needs a person (no documented handoff; the step says what to build), blocked (something stops agents outright), or needs a test (public pages cannot answer it, so a test finds out).
Try it free on your machine
The agent-ready CLI runs the same check and test on your own computer, with your own Claude Code. No account, and it sends nothing to Tanso. CLI: available now, free. Hosted: available now at app.tansohq.com (sign in).
npx @tansohq/agent-ready check yourproduct.com
- Check the public pages: the CLI reads your public pages the way an agent does, with no key or account, and writes one fix prompt per gap for a coding agent.
- Test with a real agent: your own Claude Code tries the task on your product. Then test calls your API itself with the agent's key, with no key, and with a wrong key.
- Fix it and run again: apply the fix prompts, then run the same check or test again.
Agent access
A coding agent can run it. Paste: "Read https://raw.githubusercontent.com/tansohq/agent-ready-cli/main/skills/agent-ready/SKILL.md and follow it to check yourproduct.com." Or install the plugin or skill; see the quickstart. Test starts only when the person asks. An agent can also sign in to the hosted dashboard with AgentID, as itself; approvals and payment stay with the person. Access and hosted runs: auth.md.
Want help with a fix? Contact us.