Find the bugs in your product before your users do.
Autonomous AI agents read your docs like a real developer, exercise your live API in an isolated sandbox, and surface every issue — from typos to broken workflows to security holes — ranked by severity.
Today's run
19 findings
across 47 tested workflows
Revoked API key still authorizes requests
POST /v1/keys/revoke · confirmed by 3/3 models · 99% confidence
Pagination cursor silently drops records
GET /v1/events?cursor= · reproduced 5/5 runs · 96% confidence
Rate-limit responses omit Retry-After header
429 on /v1/completions · docs claim otherwise · 91% confidence
Quickstart curl example uses a deprecated flag
docs/quickstart#step-2 · sample fails to run · 88% confidence
How it works
Set it up once. It works while you sleep.
Connect a sandbox
Point us at your public docs and drop in an API key or credentials for a sandbox / test account. Add optional guidelines on where to focus — or where to steer clear.
Agents test it deeply
A council of frontier models plans real user workflows, runs them end to end, then hammers edge cases and bounds — cross-checking each other to catch what a single model misses.
You get a ranked report
Every finding comes with severity, confidence, and a copy-paste reproduction. We tear down everything we created, so your product is left exactly as we found it.
What we find
From typos to zero-days — nothing slips through.
We test your product the way a curious, thorough developer would, then rank everything by severity so you always know what to fix first.
- Leaked secrets & auth bypasses
- Injection & access-control flaws
- Data exposed across tenants
- Broken core workflows
- Endpoints that don't match the docs
- Errors that corrupt state
- Confusing or missing error messages
- Undocumented rate limits
- Inconsistent SDK behavior
- Typos & broken links in docs
- Out-of-date code samples
- Wrong default values
Why it's different
Depth you can't get from a single model or a static scanner.
A council of models, not one
We run your tests across multiple frontier models in parallel and compare their results. More perspectives means more real issues surfaced and fewer blind spots.
Verified, not noisy
Every suspected issue is re-run and cross-checked before it reaches you. We distinguish real product bugs from doc errors and transient flakes — no daily alert fatigue.
Leaves no trace
Agents track every resource they create and tear it all down when they're done, returning your product to its exact initial state. Verified every run.
Gets smarter over time
We remember the nuances of your product across runs — so planning improves, coverage grows, and we automatically flag regressions against previous runs.
Reproducible evidence
Each finding ships with a concise summary and a copy-paste reproduction — request, response, and exact steps — so your team can confirm and fix in minutes.
Tests everything by default
No test plan required. Point us at your docs and we discover and exercise every workflow we can find — or focus exactly where your guidelines tell us to.
Always-on cadence
Runs every day. Summarized every week.
Your product changes constantly — so testing shouldn't be a one-off. TestMyDocs runs a full pass daily and delivers a clean weekly report of what's new, what regressed, and what's worth your team's attention.
- Daily unattended runs against your live product
- One digest a week — no notification fatigue
- Automatic regression detection run-over-run
1 new critical
Auth token accepted after revocation
2 high
Pagination cursor drops records on page 3
5 medium
429s return no Retry-After header
11 low
Docs: 4 broken links, 3 stale samples
Security & trust
Built to be let loose on your product — safely.
Handing an agent the keys to your API is a big ask. We designed TestMyDocs so that ask is small: scoped access, strict isolation, and a clean-up you can verify.
Sandbox credentials only
You give us a key into a test or sandbox account — never production secrets. You choose exactly what the agents can reach.
Isolated by design
Agents run in isolated sandboxes with a locked-down network allow-list, so they can only act as a real user of your product.
Guaranteed teardown
Every resource the agents create is tracked and destroyed at the end of each run, and we verify your product is back to its initial state.
Fully inspectable
Every run keeps a complete trace of what the agents did — evidence and reproduction for each finding, so nothing is a black box.
FAQ
Questions, answered.
What do you need from me to get started?+
Just three things: a link to your public docs, an API key or credentials into a sandbox or test account, and (optionally) a few notes on where to focus or what to avoid. That's it — no test scripts to write.
Do you need production access?+
No. We recommend pointing us at a sandbox or dedicated test environment. The agents only need to act as a real user of your product with the credentials you provide.
Will this leave junk in my account?+
No. Agents track every resource they create and tear it all down at the end of each run, then verify your product is back to its initial state. Leaving no trace is a core design principle.
How is this different from a normal API test suite or a security scanner?+
A test suite only checks what you already thought to write. A scanner looks for known patterns. TestMyDocs reasons about your product from your docs like a developer would — discovering workflows, probing edge cases, and using multiple frontier models to surface issues no single approach would find.
Won't AI agents flood me with false positives?+
That's exactly what we optimize against. Every suspected issue is re-verified and cross-checked across models before it reaches you, with a severity and confidence attached — so your weekly report is signal, not noise.
How often does it run and how do I hear about issues?+
It runs a full pass every day and sends a single consolidated report every week, highlighting new findings and regressions since the last run.
What kinds of products does it work with?+
Today we focus on API products with public documentation. Broader coverage is on the roadmap — join the waitlist and tell us about your product.
Ship with confidence.
Join the waitlist to be among the first teams to put an always-on council of AI testers on their product.