Sentry MCP Server
Connect AI coding agents to Sentry for live error monitoring, issues, stack traces, and AI root-cause analysis.
Updated
What is Sentry MCP?
The Sentry MCP server is a Model Context Protocol server that connects AI assistants and coding agents to your Sentry error-monitoring data — issues, events, stack traces, projects, and Seer AI analysis — so they can debug with real production context.
Sentry runs it as a hosted remote service at https://mcp.sentry.dev/mcp. Authentication uses OAuth: you log in with your existing Sentry organization and grant the agent access, so there are no API tokens to copy or rotate. The server acts as middleware over Sentry's API and is optimized for human-in-the-loop coding assistants like Claude Code and Cursor.
Beyond the hosted endpoint, Sentry also ships a self-hosted stdio option (the @sentry/mcp-server npm package) for local use or self-hosted Sentry instances, which authenticates with a Sentry user auth token instead of OAuth. The hosted remote server is the recommended path for most teams.
Tools the Sentry MCP server exposes
| Tool | What it does |
|---|---|
| search_events | AI-powered natural-language search across Sentry events, translated into Sentry's query syntax. |
| search_issues | AI-powered natural-language search across Sentry issues to find and filter problems. |
| use_sentry | Primary delegation tool used in agent mode to act against Sentry on the agent's behalf. |
Connect the Sentry MCP server
Claude Code (remote, OAuth)
- 1
Add the remote server
Run: claude mcp add --transport http sentry https://mcp.sentry.dev/mcp
- 2
Authenticate via OAuth
Start Claude Code and run /mcp, then select Sentry to open the browser OAuth flow and log in with your Sentry organization.
- 3
Verify
Ask Claude to list your Sentry projects or recent issues to confirm the connection works.
Cursor (remote, OAuth)
- 1
Open MCP settings
In Cursor go to Settings > MCP (or edit ~/.cursor/mcp.json).
- 2
Add the Sentry server
Add an entry with url set to https://mcp.sentry.dev/mcp (Streamable HTTP; clients without OAuth/HTTP fall back to SSE).
- 3
Complete OAuth
Cursor prompts you to log in to your Sentry organization; approve access, then confirm the Sentry tools appear in the MCP list.
Sample use cases
An on-call engineer asks their coding agent to triage a spiking Sentry issue
The agent pulls the issue's events and stack trace and proposes a root cause without leaving the editor.
A workflow reacts to a new high-severity Sentry issue
An agent summarizes the error and stack trace and routes a concise report to chat or a ticket.
A developer fixing a bug wants production context
The agent searches recent events and issues in natural language to confirm the failure pattern before editing code.
Security & permissions
Authentication is OAuth: you sign in through your existing Sentry organization and approve access, so no API tokens are stored client-side. The agent operates within the scopes granted at authorization and can read your Sentry issues, events, stack traces, projects, and run Seer AI analysis — treat that as access to potentially sensitive production error data. The self-hosted stdio variant instead uses a Sentry user auth token, which you scope and store yourself.
Using Sentry MCP with Gamut
In Gamut, a Sentry MCP connection lets an agent pull live error-monitoring context into automated, event- or schedule-triggered workflows — for example, on a new high-severity issue webhook an agent can fetch the issue's events and stack trace, summarize the likely root cause, and post it to the team's chat or open a tracking ticket. Gamut manages the OAuth connection and credentials, so the agent calls Sentry's tools as part of a larger workflow without you handling tokens; what it can read or change is bounded by the scopes granted during connection.
Frequently asked questions
What is the Sentry MCP server?
It's a Model Context Protocol server, hosted by Sentry at https://mcp.sentry.dev/mcp, that lets AI assistants and coding agents access your Sentry error-monitoring data — issues, events, stack traces, projects, and Seer AI analysis — to debug with real production context.
Is the Sentry MCP server free?
There's no extra charge for the MCP server itself; it's included with Sentry, which has a free Developer tier and paid plans. Note that the AI-powered search tools require you to supply your own OpenAI or Anthropic LLM provider.
How do I connect the Sentry MCP server to Claude Code or Cursor?
Point the client at the remote URL https://mcp.sentry.dev/mcp over HTTP. In Claude Code run 'claude mcp add --transport http sentry https://mcp.sentry.dev/mcp' then authenticate via /mcp; in Cursor add the URL under Settings > MCP. Both use OAuth, so you log in with your Sentry organization instead of pasting a token.
Is the Sentry MCP server hosted or self-hosted?
Both. Sentry runs a recommended hosted remote server at https://mcp.sentry.dev/mcp using OAuth, and also ships a self-hosted stdio option (the @sentry/mcp-server npm package) that authenticates with a Sentry user auth token, suitable for local use or self-hosted Sentry.