sentry

name	sentry
description	Monitor and analyze Sentry errors via MCP. Triggers: sentry, sentry issues, check errors, error tracking, show crashes.
effort	medium
context	inline

/sentry — Error Monitoring

Search, analyze, and triage Sentry errors via MCP tools. Always respond in the user's language.

Prerequisites

Sentry MCP server must be configured in .mcp.json:

{
  "mcpServers": {
    "sentry": {
      "url": "https://mcp.sentry.dev/mcp"
    }
  }
}

Or self-hosted with auth token:

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": ["-y", "@sentry/mcp-server-stdio"],
      "env": { "SENTRY_AUTH": "<TOKEN>" }
    }
  }
}

If not configured, help user set up .mcp.json first and stop.

Algorithm

1. Check MCP Availability

Verify Sentry MCP tools are available. If not — guide user through setup.

2. Determine Operation

• $ARGUMENTS = empty → show recent unresolved issues (step 3)
• $ARGUMENTS = "issues [project]" → list issues for project (step 3)
• $ARGUMENTS = "issue " → get detailed issue info (step 4)
• $ARGUMENTS = "errors [project]" → list recent error events (step 3)
• $ARGUMENTS = "analyze " → deep analysis with fix suggestion (step 5)

3. List Issues/Errors

list_project_issues(project_slug, query="is:unresolved", sort="date")

Present as compact table with events count and last seen.

4. Issue Details

get_sentry_issue(issue_id)

Show: title, stacktrace, frequency, affected users, first/last seen, tags.

5. Analyze and Fix

1. Get issue details and full stacktrace
2. Find the problematic file in local codebase:

   Grep "{error pattern}" in project files
   

3. Read the file and understand context
4. Identify root cause from stacktrace + code
5. Suggest a concrete fix with code diff
6. Offer to create a task:

   .tausik/tausik task quick "Fix: {error title}" --stack {stack} --role developer
   

6. Log Activity

If active task exists:

.tausik/tausik task log <slug> "Sentry: analyzed {issue-id}, root cause: {summary}"

Output Format

## Sentry: {project} (last 24h)

| # | Issue | Events | Users | Last Seen |
|---|-------|--------|-------|-----------|
| PROJ-456 | TypeError: null ref | 234 | 45 | 2m ago |

Rules

• Show actionable info: stacktrace, frequency, affected users
• For /sentry analyze — always find the file locally and suggest a fix
• Keep output compact — focus on what matters for fixing
• Suggest next: /debug for deep debugging, /task to start fixing, /plan to plan the fix

Gotchas

• Sentry MCP has two modes — hosted (url) and self-hosted (command). Auth tokens differ between them.
• Source maps must be uploaded for JS stacktraces to be useful — without them, line numbers point to minified code.
• Rate limiting — Sentry API has limits. Avoid fetching all issues in a loop; use search filters.