---
title: "Sentry"
description: "Control Sentry token retrieval for AI agents"
url: "https://gazebohq.com/services/sentry"
---

Error-triage agents may need Sentry API access to inspect issues, releases and project information. The Sentry preset provides an auth-token field and a link to provider settings. Gazebo stores that token and controls which agent can retrieve it; the runtime then calls Sentry directly.

## Keep error-triage access attributable

Sentry issue data can contain stack traces, request context and other sensitive application details. Select the narrowest token scopes and organizational access appropriate to the agent, and review Sentry's data-scrubbing settings separately. Gazebo's retrieval log does not show which issue the runtime later opened, nor does it remove Sentry permissions from the token. Provider records remain necessary for investigating downstream activity.

## How it works

1. Create a Sentry auth token with the minimum scopes required for your issue-triage or release workflow.
2. Choose Sentry in Gazebo and store the token in the auth_token field, keeping unrelated organizations or environments separate.
3. Grant retrieval to the relevant agent. Its runtime receives the token and makes its own authenticated Sentry requests.
4. Review credential-retrieval records in Gazebo. Revoke a delivered token in Sentry when future retrieval controls are not sufficient.

## Use cases

**Issue summaries without shared environment files:** Let a reporting runtime retrieve a dedicated Sentry token when it needs to summarize errors. Token scopes and Sentry organization membership determine the data it can access.

**Retire a temporary debugging agent:** Remove the Gazebo retrieval grant when the investigation ends. If the agent retained the token, revoke it in Sentry as well; historical retrieval records remain separate from actual issue activity.
