Auto-Triage and Fix User Feedback with Cursor Automations and Sentry MCP
Trigger a Cursor agent the moment user feedback is assigned in Sentry. It reads the issue, finds the relevant code, and opens a PR.
Before you start
SDKs & packages
- Sentry SDK installed in your app
Accounts & access
- Sentry account with a project set up
- Cursor Pro or Business subscription (required for Automations)
Knowledge
- Basic familiarity with your app's codebase and repo structure
1 Add the User Feedback widget to your app
Sentry's User Feedback widget gives users a simple way to report problems directly from your UI. When submitted, feedback appears as an issue in your Sentry project, complete with session context, tags, and optional replay links.
Add the widget using the Sentry SDK's feedbackIntegration. Configure it to match your app's look and feel, and it will attach automatically to your Sentry initialization.
import * as Sentry from "@sentry/browser";
Sentry.init({
dsn: "your-dsn-here",
integrations: [
Sentry.feedbackIntegration({
colorScheme: "system",
showBranding: false,
}),
],
}); 2 Connect the Sentry MCP to Cursor
The Sentry MCP server gives Cursor agents direct access to your Sentry issues, feedback, and debugging context. Once connected, the @Sentry tool will be available in all your Cursor agents.
- Visit mcp.sentry.dev and click Install in Cursor to configure the MCP server automatically
- Or add it manually in Cursor Settings → MCP using the server URL
https://mcp.sentry.dev/mcp - Authenticate with a Sentry auth token that has
project:readandissue:readscopes
3 Write the agent instructions
Agent instructions are the prompt your Cursor Automation runs with. They tell the agent what to do when the trigger fires: fetch the feedback from Sentry, find the relevant code, and attempt a fix. Copy the template below and adjust the bracketed placeholders for your project. You'll paste this into the Instructions field in the next step.
You are a user feedback analyst for [your-app]. You run when a Sentry User Feedback issue is assigned in the "[project]" project.
## Step 1: Gather context from Sentry
Use @Sentry MCP to fetch the assigned issue. Extract:
- The feedback message
- The page or feature the user was interacting with (from tags/context)
- Any additional context (browser, session replay link)
- Only proceed if the issue is a User Feedback type issue
## Step 2: Identify and read the relevant code
Map the feedback to a file or component in the repo:
- Use the page URL, feature name, or error context as clues
- Search the codebase if the mapping isn't obvious
Read the relevant files to understand what the user experienced.
## Step 3: Assess and fix
Determine if you can fix this with a code change. You CAN fix it if:
- It's a UI bug, copy error, or behavioral issue you can identify in the code
- You've read the relevant files and know exactly what to change
- No unverifiable assumptions are needed
- The fix is scoped to 1–3 files
If you can't fix it confidently, skip to Step 5.
## Step 4: Open a PR
- Branch: feedback/<short-description>
- Minimal edits only — match surrounding code style
- PR title: fix: <what changed> (under 70 chars)
- PR body: summarize the user feedback and what you changed
## Step 5: Comment on the Sentry issue
Use @Sentry MCP to add a comment on the issue with:
- A link to the PR (if opened)
- OR what needs changing and why it needs a human
## Rules
- Be conservative — prefer commenting in Sentry over opening a speculative PR
- Never add features or refactor surrounding code
- Scope every change to exactly what the feedback describes 4 Create a Cursor Automation
Cursor Automations are background agents that fire automatically based on a trigger, like a Sentry issue being assigned.
- In Cursor, open the Automations tab
- Click New Automation and give it a name (e.g. "Fix User Feedback")
- Under Triggers, add: *Issue assigned* → select your Sentry project
- Under Tools, enable @Sentry MCP and Open Pull Request
- Paste your instructions from Step 3 into the Instructions field
- Set Environment to *Cursor Cloud* so the agent runs without your machine being on
5 Trigger the automation with a real feedback issue
With the automation live, test it end-to-end. The agent will fetch the issue context from Sentry, read the relevant code, and either open a PR or leave a comment explaining why it passed.
- Submit feedback through your app's User Feedback widget
- In Sentry, navigate to Issues → User Feedback
- Assign the feedback issue to any team member
- Switch to Cursor and open Automations → Run History to watch the agent spin up
6 Review the agent's output
When the agent opens a PR, treat it like any other code review. For feedback the agent couldn't fix, the Sentry issue comment will explain what it found and why it passed, giving your team a head start on triaging manually.
- Check the diff to confirm the change matches the feedback description
- Read the PR body: the agent should explain what feedback it addressed and what it changed
- Run your test suite before merging
- If the fix is wrong or too broad, close the PR and tighten the agent instructions
That's it.
Your agent handles the first response.
When user feedback lands in Sentry and gets assigned, a Cursor agent investigates the root cause and opens a draft PR, so your team reviews fixes, not raw complaints.
- Configured the Sentry User Feedback widget to collect structured feedback
- Connected Sentry MCP to Cursor for in-agent issue access
- Created a Cursor Automation triggered by Sentry issue assignment
- Wrote agent instructions that triage feedback and open a fix PR
Pro tips
- 💡 Scope your Sentry project trigger narrowly: create one automation per project so each agent has the right codebase context and instructions.
- 💡 Add a validation step at the top of your instructions that checks the issue type first. If it's not User Feedback, the agent should exit early rather than wasting a run on an unrelated issue.
- 💡 Include a brief map of your repo structure or key component directories in the agent instructions so it knows where to look without searching the entire codebase on every run.
- 💡 Use Sentry's Alert Rules to automatically assign incoming User Feedback issues to a team. This way the automation fires without any manual triage step.
Common pitfalls
- ⚠️ Forgetting to grant the Sentry auth token both
project:readandissue:readscopes. The agent will fail to fetch issue details and may produce no output or a misleading error. - ⚠️ Writing agent instructions that are too broad, or not scoping the trigger to User Feedback issues. The agent will spin up for every assignment in the project and make speculative changes. Add explicit rules about what issue types it should handle and what it is and isn't allowed to touch.
- ⚠️ Not testing the full end-to-end flow before relying on the automation. The widget submission, Sentry issue creation, assignment trigger, and Cursor agent response all need to be verified together.
Frequently asked questions
What's next?
Fix it, don't observe it.
Get started with the only application monitoring platform that empowers developers to fix application problems without compromising on velocity.