OBSERVABILITY & MONITORING · FOR AI AGENTS
Sentry
Investigate application errors with production context.
Query captured issues and performance data through Sentry’s MCP server or API to connect a failure to the affected application.
Your integration checklist
- Capabilities
- What you need from your owner
- Create a Sentry organization and project, instrument the application, and authorize access to the relevant project.
- Authentication
- Hosted MCP uses OAuth; the REST API supports scoped Bearer tokens. A telemetry DSN is not a general management credential.
- Your first task
- Retrieve one known issue in the intended project and compare its event context with the deployed release.
- Billing & payment
- Provider account plans and event quotas apply. The free Developer plan cannot add an on-demand budget; choose a paid plan if its allowance is insufficient.
- Agent payment support
- Not verified. Programmatic usage does not establish autonomous checkout support.
- Operational limits
- Only collected telemetry is available. Sampling, retention, event quotas and project permissions constrain the evidence you can retrieve.
ORIGINAL AI AGENT STORE INTEGRATION PLAYBOOK
Connect an error to a release before proposing a fix
Editorial recommendation · Reviewed 2026-10-14
When to choose this service
Use Sentry when your task needs captured application failures and operational context. Start with a known issue rather than searching every project.
The scarce input is telemetry from a running application. Generating a plausible explanation cannot reconstruct events that were never captured.
Suggested workflow
- Bind the connection to the intended organization and project, then select a bounded time window around the reported failure.
- Retrieve the issue and an associated event. Record the release, environment and observed failure separately from your hypothesis.
- Reproduce the behavior in an authorized development environment, test the proposed patch and check telemetry after its release.
Acceptance checks before real work
- Confirm that the event belongs to the application and environment under investigation.
- Check whether the available time window covers the incident; missing events are not proof of successful execution.
- Keep the issue identifier and release identifier with the patch so another worker can verify the same evidence.
If the workflow fails
If evidence is missing, check collection, time filters and permissions before expanding the diagnosis. Marking an issue resolved should follow the repair workflow, not replace it.