SnapVisor

SnapVisor is a visual regression testing platform. Your CI uploads screenshots as builds, SnapVisor pixel-diffs them against an approved baseline, and your team reviews and approves the changes it finds. This connector lets Claude work that review loop: list projects and builds, pull the diffs detected in a build, approve or reject a build, dismiss a review, ignore or un-ignore a flaky change, read tests and their changes, and run the conversation on builds, tests and media. It also covers administration: projects, automation rules, members and invites, deployments, analytics and the media library.

The connector exposes that REST API through two tools: search_tools returns typed TypeScript declarations for the operations that match your request (90 operations today), and execute_typescript runs a short program that calls them. Every call goes through the same handler, OAuth scope check and plan gate as the REST API. The operation catalog is generated from SnapVisor's own OpenAPI document, so it cannot drift from the product.

SnapVisor does not generate images: no browser engine in the backend, no media-generation dependency anywhere. It compares the screenshots your own CI produced using deterministic pixel math and returns them as signed links that expire after an hour. Builds and screenshots are uploaded by your CI through project-token endpoints kept off this surface, so Claude can read and review your screenshots but cannot capture, alter or fabricate one.

Connections use OAuth 2.1 with PKCE and dynamic client registration, through a consent screen where you choose which accounts the connection covers and can narrow the scopes. Authorized applications are listed under Settings with a revoke button. MCP access requires a paid SnapVisor plan; a free account is refused. Approving a build, removing a member or deleting media are real, visible actions for your team: ask Claude to show you first and confirm before it acts. Docs: snapvisor.io/docs/agents/mcp-server.