For coding agents

Prompt AI agents to improve your software

Once your applications are instrumented with Raygun and the MCP is configured, use these prompt ideas to strengthen your software like never before.

Setup Raygun

Known failure

Known error-group resolution

Start with a Raygun error URL and give an agent the context to explain the failure, connect it to a release, and recommend a narrowly scoped fix.

Fix a known exception

Use the Raygun MCP to investigate error group [error group URL].

1. Summarize the stack trace, last seen, frequency, and affected users.
2. Correlate the group with relevant deployments and inspect a representative error instance.
3. Keep in mind that the cause of the error might not be where the error was thrown. Look back through the code to see if it's a case of a faulty value being passed through, or data being put into a bad state somewhere else entirely.
4. Before making a fix, write a test that reproduces the cause of the issue and confirm that it fails.
5. Propose the smallest code change that fixes the bug and passes the test.

Release safety

Release regression triage

Turn a suspected bad release into an evidence-backed fix plan. This prompt connects the newest deploy to new or spiking errors, real user impact, and a regression test. To use this prompt, register each deployment with Raygun as part of your deployment process.

Set up deployment tracking

Detect a release regression

Use the Raygun MCP to take a look at my [production application name]. Investigate whether the latest deployment introduced a regression.

1. Identify the latest deploy and its version.
2. Find error groups that appeared or spiked after that deploy.
3. Investigate the highest-impact groups with stack traces and full instance details.
4. Summarize affected users and sessions where available.
5. Propose the smallest code change with a regression test.

Customer escalation

Customer issue reconstruction

Begin with a customer identifier to reconstruct what they experienced across related errors, sessions, browser context, and deploy history before proposing a fix. To use this prompt, configure customer tracking in your Raygun SDK so Raygun can identify affected customers.

Set up customer tracking

Trace a customer issue

Use the Raygun MCP to investigate the production issue encountered by customer [customer email or name] in application [application name].

1. Find the customer and their related error groups and sessions.
2. Summarize the most relevant stack traces, breadcrumbs, browser/runtime context, and reproduction clues.
3. If there is enough context, explain the steps that the customer may have taken and the state of the data to reproduce this issue.
4. Propose the smallest code change with a regression test.
5. Keep in mind that the cause of the error might not be where the error was thrown. Look back through the code to see if it's a case of a faulty value being passed through, or data being put into a bad state somewhere else entirely.

Performance drop

Page performance investigation

Investigate a slow or unreliable route across traces, browser cohorts, and sessions so an AI agent can bring customer satisfaction back up.

Restore route performance

Investigate the production degradation in application [application name] on route [route or URL] using the Raygun MCP.

1. Compare the route's recent performance with its normal baseline.
2. Identify slow traces or outliers and affected browsers/environments.
3. Inspect representative sessions and breadcrumbs for reproduction context.
4. Correlate the change with deployments, code paths, or dependencies only where evidence supports it.
5. Propose the smallest code improvement with a regression or performance test.

Production eyes for coding agents

Wire Raygun MCP once.
Use AI agents every day.

Full setup