Sentry MCP Server
Sentry error, issue, and performance data for AI coding assistants.
Data last scanned today · Reviewed yesterday
Overview
This middleware connects AI development tools like Claude Code and Cursor to Sentry, enabling automated debugging and performance analysis. It acts as an interface between your AI assistant and Sentry's API to handle issues, events, and traces directly within your coding environment. The server supports both remote deployment via mcp.sentry.dev and a local stdio transport for self-hosted instances. When using natural language queries for events and issues, the server integrates with providers like OpenAI or Anthropic to process search requests before interacting with Sentry.
Our verdict
This server carries a high-risk score of 33/100 due to critical security limitations. It lacks an authentication mechanism, meaning anyone with access to the configuration can utilize the server, and it provides full write access without a read-only mode, creating a potential vector for unauthorized modifications. It is recommended only for power users who can manage the exposure risk in isolated or private development environments.
- Integrated subagent delegation for Claude Code
- Supports both remote cloud and local stdio hosting
- Maintains a small footprint with 14 direct dependencies
- No built-in authentication method
- No read-only mode preventing write actions
- License status is marked as NOASSERTION
- Not maintained by the official Sentry organization
Setup
{
"mcpServers": {
"sentry-mcp": {
"args": [
"-y",
"sentry-mcp"
],
"command": "npx"
}
}
}Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop. Some servers require API keys or tokens as environment variables — check the project's README and add an "env" object if needed.
- No read-only mode — running this grants full read/write access.
- No built-in authentication — anyone with access to this config can invoke every tool it exposes.
Tools
| Tool | Description | Risk |
|---|---|---|
| search_events | AI-powered search for Sentry events. | low |
| search_issues | AI-powered search for Sentry issues. | low |
Compatibility
| Client | Local | Docker | Remote | Read-only |
|---|---|---|---|---|
| Claude Desktop | ||||
| Cursor | ||||
| VS Code | ||||
| Windsurf | ||||
| ChatGPT |
Frequently asked questions
›What happens if I do not configure an LLM provider?
AI-powered search tools like search_events and search_issues will be unavailable, but all other MCP tools will continue to function normally.
›What permissions are required for the Sentry User Auth Token?
You need an auth token with org:read, project:read, project:write, team:read, team:write, and event:write scopes.
›How do I connect to a self-hosted Sentry instance?
You must provide the hostname via the --host flag or SENTRY_HOST environment variable, and you can add --insecure-http if your deployment does not support TLS.
›Can I disable specific server features?
Yes, you can use the --disable-skills flag or the MCP_DISABLE_SKILLS environment variable to prevent specific tools, such as Seer, from being exposed.
›What is the difference between Sentry-Bearer and Bearer tokens in remote configurations?
Bearer is reserved for OAuth access tokens, while Sentry-Bearer is used for passing an upstream API token that the server forwards without storing, validating, or refreshing.
Alternatives
Changelog
Badge
Maintain this server? Add the live badge to your README.