Sentry MCP | MCPBro

What is Sentry MCP?

Sentry MCP is an open-source MCP server designed to provide secure, structured access to your Sentry data and actions through the standardized Model Context Protocol. It allows host applications, such as Claude Desktop or IDEs, to interact programmatically with Sentry via MCP-compliant tools and endpoints. It supports both remote and local (stdio) transports, making it flexible for both cloud and self-hosted Sentry installations.

How to Configure

Production Use:
Simply access the deployed service at https://mcp.sentry.dev.

Self-Hosted Sentry (Stdio Transport):

Create a Sentry User Auth Token with appropriate scopes:

org:read
project:read
project:write
team:read
team:write
event:write

Start the MCP server:

npx @sentry/mcp-server@latest --access-token=sentry-user-token --host=sentry.example.com

Alternatively, configure with environment variables:
```
SENTRY_ACCESS_TOKEN=
SENTRY_HOST=
```

Local Development:
- Create an OAuth App in Sentry with http://localhost:5173 as the Homepage URL and Redirect URI.
- Add your credentials to .dev.vars in packages/mcp-cloudflare/:
```
SENTRY_CLIENT_ID=your_client_id
SENTRY_CLIENT_SECRET=your_client_secret
COOKIE_SECRET=my-super-secret-cookie
```
- Run the server locally:
```
pnpm dev
```

How to Use

Visit https://mcp.sentry.dev for the hosted server, or run your local/stdio instance as described above.
Use the MCP Inspector:
- Enter your server URL (e.g., http://localhost:5173) and connect.
- Authenticate as prompted (OAuth or access token).
- Discover available tools via the "List Tools" function.
- Invoke desired tool endpoints (e.g., fetch issues, list projects) directly from the Inspector or through an LLM-capable client.
For self-hosted Sentry, run with relevant tokens and connect with "stdio" mode.
Run tests with pnpm test and LLM-backed evaluation with pnpm eval (after setting OPENAI_API_KEY).

Key Features

Integration with Sentry via MCP-compliant protocol, enabling LLM/agent access to error data and operations
Supports both remote HTTP(S) and local stdio transports for flexibility across deployment types
OAuth and token-based authentication for secure access to Sentry resources
Pre-built set of discoverable, executable tools (actions/endpoints) for automation and data retrieval
Extensible and open for local development and contribution
Includes built-in Inspector tool for interactive exploration and testing

Use Cases

Allow LLM-based agents or copilots to monitor, fetch, or triage Sentry incidents within IDEs or unified desktop environments
Automate routine Sentry actions such as issue assignment, status changes, or notifications, triggered by AI workflows
Integrate Sentry observability data into custom dashboards or reporting tools powered by language models
Enable zero-touch Sentry administration or triaging via secure, standardized API without manual scripting

FAQ

Q1: Can I use Sentry MCP with self-hosted Sentry instances?
Yes, Sentry MCP supports self-hosted instances via the stdio transport mode. You'll need to provide a user authentication token with the necessary permissions and configure the host accordingly.

Q2: Do I need to build an OAuth application to test Sentry MCP locally?
For local development and testing, you'll need to register an OAuth app in your Sentry account and provide its credentials in the project’s configuration files.

Q3: Can LLMs modify my Sentry data through MCP tools?
Yes, tools exposed via MCP can both read and modify data in Sentry, depending on the tool and your token’s permissions. Always restrict provided tokens to only the required scopes.

Q4: How can I test my MCP server integrations?
Use the built-in Inspector tool by connecting to your MCP server URL. This allows you to authenticate, discover, and invoke tools interactively for easy verification.

Q5: What authentication methods are supported?
Sentry MCP supports both OAuth (for remote server) and Access Token-based (for local/stdio) authentication flows.

Tool Name	Description
listProjects	Lists all Sentry projects accessible to the authenticated user, allowing LLMs or clients to fetch project metadata and select targets for further actions.
fetchIssues	Retrieves recent or active issues (errors, exceptions) from the user's Sentry projects, optionally filtered by query parameters such as time range or status.
assignIssue	Assigns a specific Sentry issue to a user or team, enabling automated triage or workflow-driven assignment via AI agents.
listTeams	Lists all teams available within the authenticated Sentry organization, supporting collaboration or team-scoped workflows.
updateIssueStatus	Updates the status of a specific Sentry issue (e.g., resolve, ignore, unresolve) as part of agent-driven remediation or automation flows.