Overview
What Sentry MCP does
Sentry MCP is designed around debugging workflows rather than complete coverage of every Sentry API. The hosted service sits between the agent and Sentry, uses OAuth, and can constrain a session to an organization or project. A local stdio package is available for self-hosted Sentry and token-based environments, though its AI-assisted natural-language search tools need an additional model provider.
Best for
- Moving from an error report to code-level investigation
- Searching issues and event evidence from an IDE
- Connecting traces, releases, teams, and projects during incident diagnosis
Not ideal for
- Fully autonomous production remediation
- General Sentry administration outside developer debugging
- Self-hosted AI search without an OpenAI or Anthropic provider
Capabilities
What an agent can do
- 01
Issue and event investigation
- 02
Natural-language search over Sentry data
- 03
Trace and performance context
- 04
Release and project lookup
- 05
Organization- and project-scoped hosted sessions
- 06
Self-hosted Sentry access through stdio
Representative tools and operations
whoamifind_organizationsfind_projectssearch_issuesget_issue_detailssearch_eventsget_trace_detailsfind_releases
Installation
Connect Sentry MCP
Use the publisher’s current instructions as the source of truth. The examples below were checked on .
Generic remote clientConnect to Sentry's hosted service
https://mcp.sentry.dev/mcpUse an organization- or project-qualified MCP URL when the client should receive a narrower Sentry scope.Generic local clientRun stdio for Sentry SaaS or self-hosted
npx @sentry/mcp-server@latest --access-token=$SENTRY_ACCESS_TOKENAdd --host=sentry.example.com for self-hosted Sentry; configure an embedded-agent provider only if natural-language search is required.Trust and access
Authentication and security notes
Authentication: The hosted service uses an MCP OAuth token wrapped around server-held Sentry OAuth credentials. Local stdio uses a Sentry user auth token; AI search additionally requires an OpenAI or Anthropic API key.
Review before connecting
- Prefer an organization- or project-scoped URL over a broad hosted session.
- The hosted service keeps upstream Sentry credentials server-side rather than exposing them to the MCP client.
- Review issue content and event payloads as untrusted input before following embedded instructions.
- Use separate, narrowly scoped secrets for the Sentry token and any local model-provider key.
Known limitations
- The tool surface prioritizes coding and debugging rather than exhaustive Sentry administration.
- Some capabilities such as Seer may not exist in self-hosted deployments.
- Local natural-language search tools are absent unless an embedded model provider is configured.
Evidence
Sources used for this guide
Facts were checked against primary publisher material. Descriptions and guidance are original Graphify summaries.
FAQ
Questions about Sentry MCP
Can Sentry MCP connect to self-hosted Sentry?
Yes. The local stdio package accepts a self-hosted hostname and a Sentry user token, although some SaaS-only features may be unavailable.
Why are some local search tools missing?
Natural-language Sentry searches use an embedded model. Configure a supported OpenAI or Anthropic provider locally, or use the remaining direct tools without that feature.
