Monitor service health
Check Agent Studio liveness and authenticated dashboard activity without treating either signal as dependency readiness.
Use the public health endpoint to confirm that the Agent Studio web process can answer requests, then use dashboard statistics to inspect activity for the signed-in user. These signals answer different questions and do not prove that every database, provider, or integration is ready.
Before you begin
You need the deployed Agent Studio base URL. To read dashboard statistics, you must also have a valid Clerk session in Agent Studio.
200 response from /api/health proves only that the route can return its basic payload. The endpoint does not check the database, model providers, MCP servers, schedulers, or other dependencies.Check the signals
Send a GET request to the public endpoint. Replace https://studio.example.com with your deployment URL.
curl --fail-with-body --silent \
https://studio.example.com/api/healthA responding process returns JSON with status, service, and timestamp fields:
{
"status": "ok",
"service": "agent-studio",
"timestamp": "2026-09-26T10:00:00.000Z"
}The timestamp is generated for each response. Treat the value above as an example, not an expected fixed value.
Configure the probe to require an HTTP success response, status: "ok", and service: "agent-studio". Record the returned timestamp with the probe result if you need to distinguish a current response from cached output.
The observable success state is a 200 response whose JSON matches those stable values and contains a valid ISO timestamp.
Sign in to Agent Studio, then open /api/dashboard/stats on the same origin in that authenticated browser. The route requires a Clerk session; an unsigned request returns an authentication error rather than statistics.
A successful request returns this envelope:
{
"success": true,
"data": {}
}The fields inside data represent the statistics available to the current user. They can vary with workspace activity, so monitor the returned data rather than relying on the empty example object.
Add timeRange=24h, 7d, 30d, or all to the statistics URL. For example, open /api/dashboard/stats?timeRange=24h to inspect the last 24 hours.
If the parameter is absent or not one of the supported values, Agent Studio uses 7d. The success envelope remains the same while the data reflects the selected window.
Use /api/health for process liveness and dashboard statistics for authenticated activity. Add dependency-specific checks in your hosting and incident procedures when you need readiness evidence; Agent Studio does not expose those checks through /api/health.
Do not infer dependency health from low or zero dashboard counts. Those values can also mean that the signed-in user has no activity in the selected window.
Verify monitoring
Run the liveness request twice and confirm that both responses identify agent-studio and return current timestamps. In an authenticated browser, request two supported dashboard windows and confirm that both return success: true.
Troubleshooting
Next steps
Use Run Agent Studio in production to place these checks in the supported build-and-start lifecycle, or review Authentication and authorization model before automating authenticated requests.