Agentic Workflow
Connect an agent to Sentry's MCP server to inspect snapshot diffs, images, and baselines.
This feature is currently in Beta. Beta features are still in progress and may have bugs. We recognize the irony.
Connect an AI coding assistant to Sentry's MCP server so it can pull snapshot diffs, images, and baselines on your behalf. This lets you investigate snapshot failures by asking the agent, instead of clicking through the Sentry UI or downloading images by hand. It complements the local CLI workflow: MCP is built for agent-assisted investigation and review, while sentry-cli is built for local downloads and diffing.
Sentry's MCP server connects your AI assistant to Sentry over the Model Context Protocol. Follow the setup and authentication instructions there for your client (Claude Code, Cursor, VS Code, and others).
The Sentry MCP has been optimized to retrieve Snapshot results for agentic workflows. If you'd like, you can also fetch the data via the REST API and expose scripts to agents.
Connect using one of these endpoints:
https://mcp.sentry.dev/mcp: base endpointhttps://mcp.sentry.dev/mcp/{organizationSlug}: scoped to one organizationhttps://mcp.sentry.dev/mcp/{organizationSlug}/{projectSlug}: scoped to one project (recommended)
The first connection starts an OAuth flow to link your Sentry account.
For most investigations the agent needs no extra setup. By default, it can read any snapshot URL and return the diff summary plus a changed image's current, base, and diff-mask previews.
To go further, grant the agent the advanced snapshot tools by enabling the Preprod Snapshots capability on the MCP authorization screen (it's off by default; check it when you first connect, or re-authorize to add it later). These tools let the agent fetch full-resolution images, list unchanged images, and look up an app's latest baseline. They give the agent more options but take up more of its context window while enabled.
You describe what you want in natural language; the agent picks the right tool, inspects the results, and reports back. The exchanges below are representative; your agent's exact wording will vary.
Paste the snapshot URL from the failing GitHub status check:
You: This snapshot check failed:
https://my-org.sentry.io/preprod/snapshots/231949/. What changed?
Agent: 12 of 340 snapshots changed on PR #4821 (
feat/new-navvsmain), all in the navigation sidebar. The biggest diffs are:
sidebar-expanded-dark.png: 18% changedsidebar-collapsed-dark.png: 14% changedsidebar-expanded-light.png: 11% changedWant me to inspect the largest change and tell you what's different?
You: Yes, please.
Agent: [fetches and compares the current, base, and diff-mask images for
sidebar-expanded-dark.png]This one is
changedat 18% (groupSidebar/dark, fromnavigationSidebar.snapshots.tsx). The sidebar is wider and each nav item has more horizontal padding, which pushes the labels to the right. That matches the spacing work in yourfeat/new-navPR. If the extra width wasn't intended, the likely cause is the new containerwidthon the sidebar wrapper. Want me to open that component and take a look?
You: Are these changes all from my nav refactor, or did anything else regress?
Agent: All 12 changed snapshots are navigation sidebar screens, which matches your
feat/new-navPR. Nothing changed outside that component.
You: Which settings screens do we have baselines for in
web-frontend?
Agent: The latest baseline on
mainincludes three settings screens:settings-profile-light.png,settings-profile-dark.png, andsettings-billing-light.png(groupSettings).
For downloading images in bulk and local diffing, use sentry-cli. See Local Testing and the Snapshots CLI reference.
Our documentation is open source and available on GitHub. Your contributions are welcome, whether fixing a typo (drat!) or suggesting an update ("yeah, this would be better").