"Investigate a deployment" is one call that would otherwise be five — list versions, filter errors by window, group them, pull instances, read traces. Wrapping the investigation rather than the API means the agent spends its reasoning on the bug instead of on orchestration, and the token cost of the round trips disappears. The breadcrumbs and customer data are what make the difference between knowing an error happened and knowing what the person was doing when it did.
Raygun's hosted endpoint over their crash and error monitoring. The tools are written at the level of an intent rather than an API route: investigate a deployment, find related errors.
- Investigating a deployment and finding errors related to one another
- Error instance detail with stack traces
- Deployment versions, so a spike can be tied to what shipped
- Browser information, breadcrumbs and customer data on the affected sessions
Generate a personal access token in Raygun and send it as a bearer header; OAuth client registration is the alternative. Streamable HTTP, with per-client guides in their repo.
Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen