> ## Documentation Index
> Fetch the complete documentation index at: https://qodex.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Sentry and observability connectors

> Connect Sentry, Bugsnag, Datadog, Rollbar, PagerDuty, or a custom MCP server so Qodex can read production errors, reproduce them, and turn them into tests.

# Sentry and observability connectors

Connect Sentry, or another observability tool, so Qodex can read production errors and act on them: file a finding, reproduce the error, and draft a test that catches it next time.

Each connector carries a prompt you write and a run frequency. For example: "check for new production errors and reproduce each as a test."

## Supported connectors

| Connector | What Qodex reads | Credentials |
| - | - | - |
| Sentry | New issues and events from a Sentry project. | **Auth token** (an internal-integration token with `event:read` scope), **Organization slug**, **Project slug** |
| Bugsnag | Crashes and errors from a Bugsnag project. | **API token** (data access), **Project id** |
| Datadog | Monitors and events from your Datadog account. | **API key**, **Application key**, **Site** |
| Rollbar | Error occurrences from a Rollbar project. | **Read access token** |
| PagerDuty | Incidents from your PagerDuty services. | **REST API token**, **Region** |
| Custom MCP server | The tools of any remote MCP server, such as internal tooling. | **Server URL**, optional **Auth header** |

## Connect Sentry

**1. Create a Sentry token**

In Sentry, create an internal integration with the `event:read` scope and copy its token. Note your organization slug and the project slug.

**2. Open Connections**

In Qodex, open the account menu at the bottom of the left navigation, choose **Settings**, then open **Connections**. Under **Observability**, choose **Sentry**.

**3. Enter credentials**

Fill in **Auth token**, **Organization slug**, and **Project slug**, then click **Connect**. Credentials are write-only: Qodex stores them encrypted and never shows them again.

**4. Write the prompt and pick a frequency**

On the connection's page, write what each run should do, then choose a **Frequency**: **Off**, **Every 30 minutes**, **Every hour**, **Every 2 hours**, **Every 6 hours**, or **Daily**. Click **Save**.

**5. Check it works**

Click **Test connection** to confirm the credentials, or **Run now** to start a run immediately.

The other connectors follow the same steps with their own credential fields. For a custom MCP server, you can paste or upload an existing MCP `config.json` and pick the server from it.

## Use a connector in chat

Mention a connected connector by name in chat, such as `@sentry`, to use it on demand:

```text theme={null}
@sentry list the new errors from the last day and reproduce the top one as a test.
```

## Use a connector with Autopilot

The **Tests for new production errors** [Autopilot](https://qodex.ai/docs/autopilot.md) agent reads your connected error tracker every hour and drafts a test for each error seen for the first time. Every agent's page also lets you choose which connectors it may use during a run.

## Manage a connection

Open **Settings > Connections** and click the connection.

* Switch **Enabled** off to pause it. The schedule stops and Qodex cannot use its tools until you enable it again.
* Click **Replace credentials** to rotate a token.
* Click **Delete integration** and confirm **Delete** to remove it. Its schedule stops and its saved credentials are destroyed. Past run chats are kept.

If the last run failed, the page shows the error at the top.

## Related

> [card] [Autopilot](https://qodex.ai/docs/autopilot.md)
>
> Let agents turn production errors into tests.

> [card] [Findings](https://qodex.ai/docs/findings.md)
>
> Review the bugs Qodex files from production errors.

> [card] [Integrations](https://qodex.ai/docs/integrations.md)
>
> See every supported connection.

> [card] [MCP](https://qodex.ai/docs/integrations-mcp.md)
>
> Connect a coding agent to Qodex.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.