Sentry Gateway

Sentry Gateway is an MCP server that provides AI assistants with direct access to Sentry's error tracking and application monitoring platform, focused on developer-centric workflows and human-in-the-loop coding assistance. This MCP bridge enables language models to query error events, analyze stack traces, inspect issue trends, and suggest fixes by connecting to Sentry's comprehensive error telemetry through a hosted MCP endpoint.

When to Use This MCP Server

Connect this server when...

• You want AI assistance diagnosing production errors by analyzing Sentry issue details, stack traces, and breadcrumbs
• Your application uses Sentry for error tracking and you need conversational access to error trends and release health
• You are debugging a specific error and want the AI to correlate it with recent deployments, affected users, and similar events
• Your team needs automated triage of incoming Sentry issues with AI-generated severity assessments and fix suggestions
• You want to generate error summary reports for standups or incident postmortems from Sentry data

Consider alternatives when...

• Your error tracking uses Bugsnag, Rollbar, or Datadog exclusively without Sentry integration
• You only need application performance monitoring without error-level detail
• Your application does not have Sentry SDK instrumentation installed

Quick Start

Copy
# .mcp.json configuration
{
  "mcpServers": {
    "sentry": {
      "url": "https://mcp.sentry.dev/mcp"
    }
  }
}

Connection setup:

1. Ensure you have an active Sentry account with projects configured
2. Add the configuration above to your .mcp.json file
3. On first connection, complete the OAuth authentication flow in your browser
4. The server will cache your authentication for subsequent sessions

Example tool usage:

# Investigate recent errors
> Show me the top 5 unresolved issues in the backend project sorted by event count

# Analyze a specific error
> Give me the full stack trace and breadcrumbs for issue BACKEND-1234

# Review release health
> How many new errors were introduced in the latest release compared to the previous one?

Core Concepts

Concept	Purpose	Details
Issues	Error grouping	Sentry groups similar error events into issues for deduplication and tracking across occurrences
Events	Individual error instances	Each error occurrence with its own stack trace, breadcrumbs, user context, and device information
Breadcrumbs	Event timeline	Chronological trail of actions and state changes leading up to an error, providing debugging context
Releases	Version tracking	Tagged application versions that correlate errors with specific deployments for regression detection
OAuth Flow	Hosted authentication	Browser-based authentication that grants the MCP server access to your Sentry organization data

Architecture:

+------------------+       +------------------+       +------------------+
|  Sentry          |       |  Sentry MCP      |       |  AI Assistant    |
|  Platform        |<----->|  Hosted Endpoint  |<----->|  (Claude, etc.)  |
|  sentry.io       | API   |  mcp.sentry.dev  | HTTP  |                  |
+------------------+       +------------------+       +------------------+
        |
        v
+--------------------------------------------------+
|  Orgs > Projects > Issues > Events > Releases     |
+--------------------------------------------------+

Configuration

Parameter	Type	Default	Description
url	string	https://mcp.sentry.dev/mcp	Hosted Sentry MCP endpoint URL for connecting to the service
organization	string	(from OAuth)	Default Sentry organization slug for scoping operations
project	string	none	Default project slug to filter issues and events to a specific application
environment	string	production	Default environment filter for queries (production, staging, development)
time_range	string	24h	Default time window for event queries (1h, 24h, 7d, 30d)

Best Practices

1. Start with high-level issue triage before deep dives. Ask the AI to list top unresolved issues sorted by frequency or impact first. This overview helps prioritize which errors deserve detailed investigation rather than jumping into random issues that may have low user impact.

2. Use breadcrumbs for root cause analysis. When investigating a specific error, always request the breadcrumb trail along with the stack trace. Breadcrumbs show the sequence of user actions, network requests, and state changes that preceded the error, often revealing the true root cause that the stack trace alone cannot explain.

3. Correlate errors with releases for regression detection. Ask the AI to compare error rates between releases to identify regressions introduced by specific deployments. Sentry's release health data shows new, resolved, and regressed issues per release, enabling data-driven rollback decisions.

4. Filter by environment to reduce noise. Production errors are typically the highest priority. When querying issues, specify the environment filter to exclude development and staging noise. This ensures the AI focuses on errors that actually affect users rather than test-related exceptions.

5. Leverage the hosted endpoint for zero-maintenance setup. The Sentry MCP server is hosted at mcp.sentry.dev, eliminating the need for local server installation and maintenance. OAuth authentication is handled automatically, making this one of the simplest MCP servers to set up and keep running.

Common Issues

OAuth flow does not complete or times out. Ensure your browser is not blocking pop-ups from mcp.sentry.dev. Try the authentication flow in a different browser if the redirect fails. If you are behind a corporate proxy, ensure mcp.sentry.dev is whitelisted in your network configuration.

"Organization not found" when querying issues. Verify you authenticated with the correct Sentry account that has access to the organization you are trying to query. If you belong to multiple organizations, specify the organization slug explicitly in your queries rather than relying on auto-detection.

Queries return too many results or time out. Narrow your queries with specific project names, environment filters, and time ranges. Sentry can contain millions of events, and broad queries across all projects and time ranges will be slow. Start with the last 24 hours and a specific project, then expand scope only if needed.