Creativity · MCP — server

MCP Sentry Server

The MCP Sentry Server brings Sentry's error-tracking platform into the Model Context Protocol. An MCP client like Claude Desktop or Cursor can list unresolved issues, fetch a stack trace with source context, follow the release and commit it regressed on, and open a PR — closing the loop between production errors and code changes.

MCP facts

Kind: server
Ecosystem: anthropic-mcp
Language: TypeScript / Node.js
Transports: stdio, http

Capabilities

Tools: list_issues, get_issue_details, get_event (with full stack trace)
Tools: assign_issue, resolve_issue, list_releases, get_release_deploys
Resources: sentry://issue/<id> — full issue with breadcrumbs and tags
Auth: OAuth via hosted server, or auth token for stdio/local use

Install

npx -y @sentry/mcp-server

Configuration

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": ["-y", "@sentry/mcp-server"],
      "env": {
        "SENTRY_AUTH_TOKEN": "sntryu_xxx",
        "SENTRY_ORG": "your-org"
      }
    }
  }
}

Frequently asked questions

Should I use the hosted Sentry MCP or the local one?

For personal use the stdio server with an auth token is fastest. For teams prefer mcp.sentry.dev: it speaks OAuth, so each engineer's permissions map directly to their Sentry role, and audit logs live inside Sentry.

What can the agent actually do with Sentry data?

A typical workflow: list top unresolved issues for a release, pull the full stack trace for the worst one, cross-reference the failing commit in GitHub MCP, and open a PR with the fix — all in a single Claude conversation.

Is sensitive data in events exposed to the model?

Yes — breadcrumbs and tags may include PII. Configure Sentry's data-scrubbing rules before connecting an LLM, and restrict the token's project scope to what the agent actually needs.

Sources

Sentry MCP server (official) — accessed 2026-04-20
Sentry auth tokens — accessed 2026-04-20