Sentry MCP — Use Cases, Install & Live Demo

Why use it

Key features

First-party — published by Sentry, follows their auth model
Filter issues by release, environment, user impact, age
Full stacktrace + breadcrumbs for any event
Read-only by default — won't accidentally resolve issues

Live Demo

What it looks like in practice

sentry.replay ▶ ready

0/0

Install

Pick your client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": [
        "-y",
        "@sentry/mcp-server"
      ]
    }
  }
}

Open Claude Desktop → Settings → Developer → Edit Config. Restart after saving.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": [
        "-y",
        "@sentry/mcp-server"
      ]
    }
  }
}

Cursor uses the same mcpServers schema as Claude Desktop. Project config wins over global.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": [
        "-y",
        "@sentry/mcp-server"
      ]
    }
  }
}

Click the MCP Servers icon in the Cline sidebar, then "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": [
        "-y",
        "@sentry/mcp-server"
      ]
    }
  }
}

Same shape as Claude Desktop. Restart Windsurf to pick up changes.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "sentry",
      "command": "npx",
      "args": [
        "-y",
        "@sentry/mcp-server"
      ]
    }
  ]
}

Continue uses an array of server objects rather than a map.

~/.config/zed/settings.json

{
  "context_servers": {
    "sentry": {
      "command": {
        "path": "npx",
        "args": [
          "-y",
          "@sentry/mcp-server"
        ]
      }
    }
  }
}

Add to context_servers. Zed hot-reloads on save.

claude mcp add sentry -- npx -y @sentry/mcp-server

One-liner. Verify with claude mcp list. Remove with claude mcp remove.

Use Cases

Real-world ways to use Sentry

Triage a fresh production incident in 5 minutes

👤 On-call engineers ⏱ ~5 min intermediate

When to use: PagerDuty just woke you up. Sentry says errors are spiking. You need to know what, why, and whether to revert — fast.

Prerequisites

Sentry org slug + project slug — Look at any Sentry URL: sentry.io/organizations/<ORG>/issues/?project=<ID>
Sentry user auth token with event:read and project:read — sentry.io/settings/account/api/auth-tokens/

Flow

Find the top NEW issue in the last hour

What's the top new issue in our web-prod project in the last hour, ranked by event count?✓ Copied

→ Single issue with title, event count, users affected, first seen timestamp
Pull the latest event with full stacktrace + breadcrumbs

Get the latest event for that issue. Show me the stacktrace, the release, and the last 5 breadcrumbs before the crash.✓ Copied

→ File:line of the throwing function + sequence of user actions before the error
Identify the introducing release

Was this issue first seen in the same release as it appeared, or did it carry over? Compare the release tag.✓ Copied

→ Yes/no with confidence — drives revert decision

Outcome: A 3-line incident summary you can paste into Slack: what's broken, who's affected, which release caused it, recommended action.

Pitfalls

If your release tags aren't wired up, you can't tell which deploy introduced the bug — Set up sentry-cli releases in your CI before relying on this — without it, you're guessing
Stacktrace is in minified JS and unreadable — Verify sourcemaps are uploaded — sentry-cli sourcemaps upload should be in your build pipeline

Combine with: github · linear

Cross-reference Sentry errors with the GitHub commits that caused them

👤 Senior engineers debugging recurring issues ⏱ ~30 min advanced

When to use: An error keeps coming back after each release. You suspect a specific code path but want to confirm.

Prerequisites

Both Sentry MCP and GitHub MCP installed — See the github guide for setup
Releases tagged with git SHA in Sentry — Use sentry-cli releases new $SHA in CI

Flow

List the issue's history across releases

For Sentry issue WEB-3a91, list every release where it appeared and the event count per release.✓ Copied

→ Table showing the issue spiking after specific deploys
For each spike, fetch the diff

For the 3 releases with the highest event count, use the GitHub MCP to get the commit diff. What files did each release change?✓ Copied

→ Common files across the spikes — the smoking gun
Form a root-cause hypothesis

Based on those diffs, what's the most likely root cause? Be specific — point to the line(s).✓ Copied

→ Line-level theory you can validate by checking the code

Outcome: A specific code-level hypothesis with evidence from both Sentry's events and GitHub's diffs.

Pitfalls

Multiple commits in a release — hard to isolate which one is responsible — Use git bisect style: deploy a build with only half the commits and check if the error rate drops

Combine with: github

Generate a weekly engineering quality report from Sentry data

👤 Engineering managers ⏱ ~20 min intermediate

When to use: Friday afternoon, before next week's planning. You want to know if quality is trending up or down.

Prerequisites

Read access to all your projects in Sentry — Token scoped to org:read + project:read + event:read

Flow

Pull error and crash-free session counts for the week

For our org, give me crash-free session % per project for this week vs last week. Flag any project where it dropped.✓ Copied

→ Per-project comparison with deltas
Identify the top contributors to error volume

What 5 issues are responsible for the most events this week, across all projects?✓ Copied

→ Concrete issue list with event counts and links
Suggest focus for next week

Based on this data, what should the team prioritize fixing next week? Consider both volume and user impact.✓ Copied

→ 3 prioritized recommendations with reasoning

Outcome: A 1-page report you can share in your weekly engineering review with concrete priorities.

Pitfalls

Volume metrics are dominated by one noisy issue, masking everything else — Filter that issue out and re-rank — sometimes the noisiest isn't the most important

Combine with: linear · notion

Quantify the user impact of a known bug before deciding to fix

👤 Product managers, tech leads triaging the bug backlog ⏱ ~15 min beginner

When to use: There's a known bug and the team is debating priority. You need data, not opinions.

Flow

Pull the issue's user impact stats

For Sentry issue WEB-3a91, how many unique users have hit it in the last 30 days, and what % of total active users is that?✓ Copied

→ Absolute count + percentage
Segment the affected users

Among affected users, what's the distribution by browser, OS, and account tier (paid vs free)?✓ Copied

→ Breakdown that reveals if it's an edge-case or core-flow issue
Compare to other open issues

Rank our top 10 open bugs by users-affected this month. Where does this one sit?✓ Copied

→ Context for prioritization

Outcome: A data-driven priority recommendation: fix now, fix this sprint, or defer.

Combinations

Pair with other MCPs for X10 leverage

sentry + github

Sentry → identify bad release → GitHub → find the commit that introduced it → draft revert PR

Issue WEB-3a91 spiked after release [email protected]. Find the commits in that release on GitHub, identify the most likely culprit, and draft a revert PR.✓ Copied

sentry + linear

Auto-create Linear issues from new Sentry issues exceeding an event threshold

Find any new Sentry issues from the last 24h with >100 events. For each, create a Linear bug ticket assigned to the on-call engineer.✓ Copied

sentry + notion

Weekly engineering quality report posted to Notion

Pull this week's Sentry stats and create a Notion page in the Engineering / Weekly Reports database.✓ Copied

Tools

What this MCP exposes

Tool	Inputs	When to call	Cost
list_issues	organization, project, query?, sort?, limit?	Search the issue stream by status/release/environment/age	1 API call
get_issue	issue_id	Get full metadata for a specific issue including release range	1 API call
get_event	issue_id, event_id?	Pull the actual stacktrace and breadcrumbs for debugging	1 API call
list_releases	organization, project	See deploy timeline and which release introduced what	1 API call
list_projects	organization	Discover what projects exist in your org	1 API call

Cost & Limits

What this costs to run

API quota: Sentry: 40 req/sec per token (very generous). No daily cap on the API itself.
Tokens per call: 200–1000 tokens per issue/event response; large stacktraces can hit 5k
Monetary: Free tier: 5k errors/month. Paid plans bill on event volume, not API usage.
Tip: API is free; what you're really paying for is event ingestion. Use sampling rules in Sentry itself to keep ingestion costs predictable.

Security

Permissions, secrets, blast radius

Minimum scopes: org:read project:read event:read

Credential storage: Sentry user auth token in env var SENTRY_AUTH_TOKEN

Data egress: All calls to your Sentry instance (sentry.io or self-hosted)

Never grant: org:write project:admin member:write

Troubleshooting

Common errors and fixes

401 Invalid token

Token expired or doesn't have the requested scopes. Re-create at sentry.io/settings/account/api/auth-tokens/

Verify: curl -H "Authorization: Bearer $SENTRY_AUTH_TOKEN" https://sentry.io/api/0/organizations/

404 Project not found

The project slug is case-sensitive and must match the URL exactly. Check sentry.io/settings/projects/

Empty stacktrace

Sourcemaps not uploaded. Run sentry-cli sourcemaps upload as part of your build pipeline.

Alternatives

Sentry vs others

Alternative	When to use it instead	Tradeoff
Datadog APM MCP	You're already on Datadog and want unified APM + errors	More expensive, less focused on errors specifically
Rollbar / Bugsnag MCP	You already pay for them	Smaller community-built MCP ecosystem

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills