M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
S

Sentry MCP Server

Official
Sentry's official MCP server for interacting with Sentry via LLMs.
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #121 of 438
Stars
★ 927 Very popular
Transport
stdio (local process) · Streamable HTTP
Runtime
Hosted (no local runtime) · Node.js
Credentials
API key / credential required
License
Other / unspecified
Last commit
60FMRS · C

Sentry's official MCP server focused on developer debugging scenarios, interacting with Sentry API via remote or stdio transports. Note permissions and external LLM dependencies.

Strongest · Documentation 14/20 Weakest · Reliability 9/20

Reliability
9/20
Security and permissions
14/20
Maintenance
10/20
Documentation
14/20
Setup experience
13/20
Why each score
Reliability 9/20
The README describes a coherent architecture (remote Cloudflare transport plus stdio) covering init, OAuth, skills-to-tools mapping, and controllable dependencies such as --host, --insecure-http, and --disable-skills, with error paths at least partially diagnosable. However, this is a static review: no CI workflows, test files, or lockfiles were supplied, and the README itself is untrusted evidence. The provided server manifest is only OpenTelemetry semantic-convention attributes (server.address / server.port) and says nothing about actual server behavior. The 93 open issues and the statement that AI-powered tools silently disappear without an LLM provider mean the exposed tool list is environment-dependent and behavior is inconsistent across configurations. Because execution is unverified, reliability is held below 12 and conservatively scored 9.
Security and permissions 14/20
No red-line issues: examples use placeholder tokens ("your-token", "sk-...") rather than real secrets, and there is no evidence of malware, credential theft, covert exfiltration, or irreversible destructive defaults. Credential handling is comparatively thoughtful — the remote design separates "Sentry-Bearer" (upstream token, never stored/validated/exchanged/refreshed by the worker) from "Bearer" (MCP OAuth token), which is good data-flow disclosure. Least privilege is partially addressed: required token scopes are documented (org:read, project:read, project:write, team:read, team:write, event:write) and the exposed skill set can be narrowed (?skills= / ?disable-skills=). Deductions: the documented default scopes include write permissions (project:write, team:write, event:write) even though the service is framed as a debugging/read-oriented assistant; no confirmation flow for side-effectful operations is described; and the exact data boundary for third-party LLM providers (OpenAI/Anthropic/OpenRouter) used by AI-search tools is not fully specified in the supplied source. Scored 14.
Maintenance 10/20
The repository is not archived and shows active development signals: a production deployment, named npm release channels (@sentry/mcp-server and an experimental variant), contributor documentation (CLAUDE.md/AGENTS.md), and a docs/ tree, so versioning/release paths exist. Deductions: the license is NOASSERTION (no license could be asserted), which is a serious governance gap; 93 open issues are listed but issue-response quality/triage is unverifiable; and there is no evidence of a security-response channel or dependency-update cadence. Because the prerequisites for full marks — a clear license and a security-response path — are absent, the conservative midpoint-low score is 10.
Documentation 14/20
The README is layered and practical: install via Claude plugin and npx, remote-header and stdio JSON config examples, both SaaS and self-hosted variants (--host, --insecure-http), a completed environment-variable table, token scope list, and explicit limitations (AI tools require an LLM provider, Seer may be unavailable on self-hosted, provider auto-detection is deprecated). It also references docs/testing/*.md and docs/operations/embedded-agents.md. Deductions: per-tool parameter documentation, rate limits, and costs are not present in the supplied source and are deferred to referenced docs that cannot be verified; troubleshooting is promised (CLAUDE.md, testing playbooks) but not evidenced here, so some hidden assumptions remain (e.g., why AI tools are missing). Scored 14.
Setup experience 13/20
Setup is a short, concrete path with working client-config examples: a one-line npx command with --access-token for stdio, a JSON mcpServers block with custom headers for remote clients, and a Claude Code plugin install command. Self-hosted cases are explicitly handled (--host, --insecure-http, --disable-skills). Deductions: per the static calibration rule, setup cannot exceed 15 without verifiable execution evidence (real CI plus committed tests), and the supplied material (README plus manifest) contains neither. In addition, the default tool set silently depends on an LLM provider whose selection now requires an explicit EMBEDDED_AGENT_PROVIDER variable, adding a hidden prerequisite to achieve the documented full tool list. Scored 13.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessUses the networkChanges third-party account data

Best for

  • Developers using coding agents like Claude Code and Cursor
  • Workflows requiring interaction with Sentry for debugging

Not for

  • Not suitable for general-purpose MCP servers needing full Sentry functionality
  • Not suitable for users without a Sentry account or self-hosted instance

Required permissions

  • Requires a Sentry User Auth Token with scopes: org:read, project:read, project:write, team:read, team:write, event:write

Risks and side effects

  • AI-powered search tools rely on external LLM providers, potentially exposing data to third parties
  • Automated code review tools may produce false positives and should not be blocking requirements

Setup

Before you start

Runtime：Hosted (no local runtime) · Node.js

SENTRY_ACCESS_TOKEN requiredsecret Required Sentry User Auth Token created in Sentry settings with scopes like org:read, project:read/write, team:read/write, event:write.
OPENAI_API_KEY optionalsecret OpenAI API key, needed for AI-powered search tools when using OpenAI; get it from platform.openai.com.
ANTHROPIC_API_KEY optionalsecret Anthropic API key, needed for AI-powered search tools when using Anthropic.
OPENROUTER_API_KEY optionalsecret OpenRouter API key, needed for AI-powered search tools when using OpenRouter.
Other optional settings (5)
EMBEDDED_AGENT_PROVIDER optional Explicitly select the LLM provider: 'openai', 'azure-openai', 'anthropic', or 'openrouter'; auto-detection is deprecated.
OPENROUTER_MODEL optional Optional OpenRouter model, defaults to 'openai/gpt-5.6-luna'.
OPENROUTER_REASONING_EFFORT optional Optional OpenRouter reasoning effort, defaults to 'high'.
SENTRY_HOST optional Hostname for self-hosted Sentry (e.g. sentry.example.com); if unset the CLI targets Sentry SaaS.
MCP_DISABLE_SKILLS optional Comma-separated list of skills to disable (e.g. 'seer') to hide unsupported tools.
  1. Ensure Node.js is installed.
  2. Run the server using npx, providing a Sentry User Auth Token via command-line argument or environment variable.
  3. For remote use, configure MCP clients to point to https://mcp.sentry.dev/mcp.
  4. For AI-powered search tools, configure an LLM provider (e.g., OpenAI) with API key.
.cursor/mcp.json 
{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": [
        "@sentry/mcp-server"
      ],
      "env": {
        "SENTRY_ACCESS_TOKEN": "your-token",
        "EMBEDDED_AGENT_PROVIDER": "openai",
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

Shown for Cursor. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "sentry": {
      "command": "npx",
      "args": [
        "@sentry/mcp-server"
      ],
      "env": {
        "SENTRY_ACCESS_TOKEN": "your-token",
        "EMBEDDED_AGENT_PROVIDER": "openai",
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add sentry -e SENTRY_ACCESS_TOKEN=your-token -e EMBEDDED_AGENT_PROVIDER=openai -e OPENAI_API_KEY=sk-... -- npx @sentry/mcp-server

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

Connect via stdio (npx @sentry/mcp-server) or remote (https://mcp.sentry.dev/mcp), then check that Sentry tools appear in the client's tool list; alternatively open MCP Inspector, enter the server URL, and connect to trigger the OAuth flow and list tools.

Troubleshooting

  1. Check SENTRY_ACCESS_TOKEN is set and valid
  2. For AI search tools, confirm EMBEDDED_AGENT_PROVIDER and corresponding API key are set
  3. For self-hosted, verify --host parameter is correct, add --insecure-http if needed
  4. If OAuth flow has issues in MCP Inspector, try using localhost instead of 127.0.0.1

Things to try

Once connected, you can ask your AI assistant things like:

  • What projects and teams do I have in Sentry?
  • Resolve and triage this Sentry issue for me
  • Search my recent error events in natural language (search_events)

Use cases

Debugging Sentry errors and exceptions
Viewing and analyzing issues and traces
Querying events and performance data

Supported clients

Claude
Cursor

Listed from the project's documentation, not tested by this site.

Overview

Sentry's MCP service is primarily designed for human-in-the-loop coding agents. Our tool selection and priorities are focused on developer workflows and debugging use cases, rather than providing a general-purpose MCP server for all Sentry functionality. This remote MCP server acts as middleware to the upstream Sentry API, optimized for coding assistants like Cursor, Claude Code, and similar development tools. It's based on Cloudflare's work towards remote MCPs.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 62.9k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 391fdb4ca36a Data synced 2026-10-12 Read the FMRS scoring method