1 · Create an API key
Dashboard → API keys → Create key. Keys are available on every plan. The key is shown once — copy it then. Only a hash of it is stored, and revoking it deletes it immediately. Checks started with the key belong to your team and spend its balance like any other.
2 · Add CheckMyApp to your agent
Claude Code — one line in your terminal:
claude mcp add --transport http checkmyapp https://checkmyapp.dev/mcp \
--header "Authorization: Bearer cma_YOUR_KEY"Cursor — add this to ~/.cursor/mcp.json (or .cursor/mcp.json in one project) and set CHECKMYAPP_API_KEY in your environment, so the key never lands in a repository:
{
"mcpServers": {
"checkmyapp": {
"url": "https://checkmyapp.dev/mcp",
"headers": {
"Authorization": "Bearer ${env:CHECKMYAPP_API_KEY}"
}
}
}
}Any other MCP client — connect to https://checkmyapp.dev/mcp over Streamable HTTP with the header Authorization: Bearer cma_….
3 · Ask for what you want
You talk to your agent as usual; it picks the tools. For example:
Add my app https://app.example.com to CheckMyApp with the test account qa@example.com — the password is QA_PASSWORD in .env.test. Scenarios: a signed-in user can create an invoice and download it as a PDF; search finds an invoice by number. Then run a check and fix what it finds.
I just deployed. Run a CheckMyApp check on app.example.com, wait for it, and work through the review.
What did CheckMyApp find on my apps since yesterday?
Turn on Daily Watch for app.example.com.
A check takes about 20–40 minutes; your agent can wait for it or come back to it.
What your agent can do
list_appsThe apps your team has added.create_appAdd an app: its address, scenarios, limits, notes, and its test accounts — one, or several named ones such as “admin” and “free user”.update_appChange an app’s scenarios, limits, notes or test accounts.start_checkCheck an app you added, or any address.get_check_statusWhere a running check is, without waiting.wait_for_runWait for a check to finish and return its verdict.wait_for_reviewWait for a check to finish and return its findings, ready to fix.get_reviewThe findings in the shape you fix from: each symptom, its evidence, and how to know it is gone.get_verdictThe verdict of a check, or the latest one for an address.latest_resultsEvery app’s last check and what is new since the one before.enable_watchTurn on Daily Watch for an app, within your plan.disable_watchTurn Daily Watch off.What your plan allows
Every tool above works on every plan, Free included. A plan is a balance, and every check — scheduled, started by your agent or from the dashboard — spends it at its own price:
- Free$3.00 of checks, once, for the whole team; a check typically costs $0.72–$2.25, and one that finds nothing changed a few cents. 1 watched app, checked daily, on a 7-day trial; top up from $10 any time.
- Starter$29.00 of checks every month, spent on anything — recurring checks, your agent, the dashboard; a check typically costs $0.72–$2.25, and one that finds nothing changed a few cents. Any number of watched apps, daily or every 6 hours; top up from $10 any time.
- Growth$99.00 of checks every month, spent on anything — recurring checks, your agent, the dashboard; a check typically costs $0.60–$1.88, and one that finds nothing changed a few cents. Any number of watched apps, daily or every 6 hours; top up from $10 any time.
- Business$499.00 of checks every month, spent on anything — recurring checks, your agent, the dashboard; a check typically costs $0.48–$1.50, and one that finds nothing changed a few cents. Any number of watched apps, daily or every 6 hours; top up from $10 any time.
Your agent is told your balance and what a check of each app usually costs when it connects, and says so before it spends the last of it. Every finished check comes back with its price and the work it paid for. When the balance runs out, the answer comes with two links: top up, or upgrade. Plans are on Pricing.
how to read what comes back
- Each finding names a symptom and the sentence that means it is gone — not a file or a fix. How to fix it is your agent’s call, with your code in front of it.
- After a fix, start another check: the next verdict says whether the symptom is gone.
- Test accounts and scenarios are covered in Checking pages behind a login and Your own scenarios.