Sentry MCP
Investigate Sentry issues, events, projects, and application health.
Verdict
Common use cases
- Investigate spike in production errors
- Summarize release health before deploy
- Compare error rates across environments
- Identify top issues by user impact
- Pull stack traces for debugging sessions
Integration
- Vendor
- Sentry MCP
- Category
- other
- Auth
- OAUTH2
- Composio slug
sentry_mcp
Tools
Setup
Setup guide
- 11. Open your Switchy workspace and navigate to Settings > Integrations. 2. Find Sentry MCP in the list and click Connect. 3. You'll be redirected to Sentry's OAuth consent screen—log in with your Sentry account if prompted. 4. Review the requested scopes (usually org:read, project:read, event:read) and click Authorize. 5. Switchy confirms the connection and returns you to the integrations page. 6. Open any Project and type '@Sentry' in the chat to verify the connection—you should see the MCP respond to queries. 7. Test it by asking '@Sentry show me the top 5 errors from the last 24 hours' or '@Sentry what's the error rate for the mobile app this week'. 8. If the MCP doesn't respond, check that your Sentry organization has active projects with recent events—empty orgs won't return useful data.
What teammates see: by default, memories from Sentry MCP are scoped to the Project (PROJECT visibility) - you can mark any memory PRIVATE or share it ORG-wide.
Works well with
Top models
Compatibility data appears once enough Projects have used this MCP together with a given model.
How Switchy teams use it
Starter prompts
Top Errors This Week
@Sentry show me the top 10 errors from the last 7 days, grouped by issue title and event count.
Release Health Check
@Sentry compare error rates and crash statistics between the last two releases in our production environment.
User Impact Report
@Sentry list issues from the past 48 hours sorted by number of affected users, include error message and project.
Performance Bottlenecks
@Sentry find the slowest transactions in the last 24 hours, show me p95 latency and throughput for each.
Stack Trace Deep Dive
@Sentry get the most recent stack trace and event details for issue ID 12345, include breadcrumbs and tags.
Example outputs
Illustrative - representative of the model's voice and quality, not literal recordings.
@Sentry show me all unresolved errors from the last 24 hours in our production environment, grouped by frequency
Here's what I found in your production environment over the last 24 hours: 1. **TypeError: Cannot read property 'id' of undefined** (checkout-flow.js:47) — 127 occurrences, affecting 89 users. First seen 18 hours ago. 2. **NetworkError: Failed to fetch** (api-client.js:203) — 43 occurrences, 31 users. Intermittent since 6 hours ago. 3. **ReferenceError: analytics is not defined** (tracking.js:12) — 19 occurrences, 12 users. Started 3 hours ago. All three are unresolved. The checkout TypeError is spiking — you may want to prioritize that investigation.
This example shows error aggregation and triage — Sentry's core strength. Requires read access to your Sentry projects. The MCP returns structured error data; the AI summarizes it for quick scanning. Real output depends on your project's current error state and retention settings.
@Sentry create an issue in our backlog for that checkout TypeError and assign it to the frontend team
I've created issue **FRONT-284** in your Sentry project and linked it to the checkout TypeError. The issue is assigned to the frontend team and includes: - Stack trace from the most recent occurrence - Affected user count (89) - Browser distribution (Chrome 78%, Safari 15%, Firefox 7%) - Link to the Sentry event stream The issue is now visible in your Sentry dashboard and any connected project management tools you've configured.
Demonstrates write capability — turning errors into tracked work items. Requires write permissions in Sentry. The MCP can create issues but can't guarantee assignment if your team structure or integrations aren't configured. Side-effect: this action is immediate and creates a permanent record.
@Sentry compare error rates between our staging and production environments for the last week and tell me if we're catching issues before release
Over the last 7 days: **Staging:** 34 unique errors, 412 total events. Top issue: database timeout (staging-db latency). **Production:** 18 unique errors, 1,847 total events. Top issue: the checkout TypeError we discussed earlier. Only 3 of the 18 production errors appeared in staging first. That's a 17% catch rate — most production errors are surfacing for the first time in prod. The checkout TypeError never triggered in staging, likely because staging traffic doesn't exercise that code path under load.
This synthesis example pairs Sentry's multi-environment data with the AI's comparative reasoning. Useful for release confidence checks. Accuracy depends on consistent error tracking across environments and comparable traffic patterns. The AI interprets trends; you'll need to verify the underlying event data.
Use-case deep-dives
When Sentry MCP speeds up incident retrospectives
A 6-person engineering team ships a bug that crashes the checkout flow on Friday afternoon. By Monday standup, the on-call engineer needs to pull error traces, identify the commit that introduced the regression, and draft a postmortem. Sentry MCP is the right call here because OAuth2 auth means the AI can query error groups, stack traces, and release metadata without copy-pasting tokens or screenshots into chat. The team saves 20 minutes per incident by letting the AI correlate error spikes with deploy timestamps. The threshold: if your team triages fewer than 2 incidents per month, the OAuth setup overhead isn't worth it—just paste the Sentry URL into chat. But if you're debugging production issues weekly, this MCP turns Switchy into a faster incident command center.
Why support teams use this MCP for error lookups
A 3-person support team at a B2B SaaS company gets a ticket: 'The export feature is broken for our account.' The support rep needs to check if Sentry logged errors for that customer's org ID in the past 48 hours. Sentry MCP wins here because OAuth2 scopes let the AI search issues by custom tags (like customer ID or environment) without exposing the full error database to every support agent. The rep asks 'any export errors for account 4729 since Tuesday' and gets a yes/no answer in 10 seconds. The trade-off: if your support team doesn't tag errors with customer metadata in Sentry, the MCP can't filter effectively—you're better off just linking the Sentry dashboard. But if you tag consistently, this MCP cuts escalation research time in half.
When this MCP helps product teams prioritize bug fixes
A 5-person product team starts sprint planning and needs to decide whether to tackle the 'image upload flakiness' bug or ship a new feature. The PM asks the AI to pull the past two weeks of error volume and affected user count for the upload flow. Sentry MCP is the right tool because OAuth2 access lets the AI query aggregated stats (error frequency, unique users, browser breakdown) without the PM needing a Sentry login or knowing the query syntax. The team sees 340 users hit the error last week, up from 80 the week before, and bumps the bug to P0. The boundary: if your team already has a Sentry dashboard pinned in Slack, the MCP adds minimal value—you're just moving the graph into chat. But if your PM or designer doesn't live in Sentry, this MCP surfaces error impact during planning without context-switching.
Frequently asked
What does Sentry MCP do in Switchy?
Sentry MCP connects your Sentry error tracking to Switchy's AI workspace, letting your team query crash reports, stack traces, and issue trends through natural language. Instead of switching between Sentry's dashboard and your chat, you ask questions like 'show me production errors from the last hour' and get structured answers. It's useful when debugging incidents or reviewing release health without leaving your workflow.
Do I need admin access to connect Sentry MCP?
You need a Sentry account with OAuth permissions to authorize the connection. Sentry uses OAuth2, so you'll authenticate through their standard flow and grant Switchy read access to your projects and issues. You don't necessarily need org admin rights, but you'll only see data from projects your Sentry account can access. Check with your Sentry admin if you're unsure about your permissions.
Can Sentry MCP resolve issues or assign bugs to teammates?
Not yet. The current integration focuses on read access — querying errors, viewing stack traces, and analyzing trends. You can't mark issues as resolved, change assignees, or modify tags through Switchy. For those actions, you still need to open Sentry directly. Think of it as a reporting layer, not a full control panel for incident management.
How is this different from just opening Sentry in a browser tab?
Sentry's UI is built for deep investigation; Switchy's MCP is built for quick answers during conversations. You can ask 'which endpoint is throwing 500s today' without navigating filters and dashboards. It's faster for triage and status checks, but you'll still use Sentry's interface when you need to drill into individual events or configure alerting rules.
Who on the team should connect Sentry MCP to Switchy?
Whoever owns your team's Sentry account and has visibility into the projects you want to query. Usually that's a backend engineer or DevOps lead. Once connected, anyone in your Switchy workspace can ask questions about the errors that account can see. If you have multiple Sentry orgs, you might connect more than one account to cover different environments.