SnapVisor MCP Server for Claude and ChatGPT
By Devino Solutions
Overview
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.
How to connect
- Claude: Open the SnapVisor listing in the Claude connectors directory and choose Connect. Claude listing SnapVisor exposes 2 tools here.
- ChatGPT: Open the SnapVisor listing in the ChatGPT app directory and choose Connect. ChatGPT listing SnapVisor exposes 24 tools here.
Tools
createBuildComment
ChatGPTthreadId, optionally anchor the comment to a screenshot diff, or attach it to your pending review with addToReview. Use it when the user wants to leave feedback on a build or on a specific screenshot diff. Requires the OAuth scope(s): comments:write (personal access tokens are not scope-restricted).createProject
ChatGPTcreateReview
ChatGPTAPPROVE to accept all changes as the new baseline, or REJECT to flag them for rework. Optionally pass per-screenshot decisions in snapshots. Confirm with the user first: the decision is reported to GitHub or GitLab, where it can unblock or block a merge. Use it after reviewing diffs to approve or reject a build. Requires the OAuth scope(s): reviews:write (personal access tokens are not scope-restricted).deleteBuildComment
ChatGPTdismissReview
ChatGPTgetAccountAnalytics
ChatGPTgetBuild
ChatGPTstatus is pending or progress (processing is still running). It then settles on no-changes (nothing to review), changes-detected (diffs found — review needed), or error, aborted or expired. On changes-detected, call listBuildDiffs to inspect the individual screenshot diffs. A reviewed build reads accepted or rejected. Requires the OAuth scope(s): projects:read (personal access tokens are not scope-restricted).getBuildComment
ChatGPTgetMe
ChatGPTgetMe first to discover which accounts (the slug field) the token can act on — slug is the {owner} path parameter used by every project-scoped endpoint. Then call listProjects with that owner to find the projects you can work with. Requires the OAuth scope(s): profile (personal access tokens are not scope-restricted).getProject
ChatGPTignoreChange
ChatGPTlistBuildComments
ChatGPTlistBuildDiffs
ChatGPTneedsReview to return only the diffs that require review. Use it to see what visually changed in a build before reviewing it. Agent workflow: call this after getBuild reports changes-detected. Inspect the diffs, then call createReview with APPROVE or REJECT. Requires the OAuth scope(s): projects:read (personal access tokens are not scope-restricted).listBuilds
ChatGPTsearch to match builds by name, branch or commit, and distinctName to return only the latest build per name and commit. Use it to find a build by branch, commit or name, or to see the latest builds of a project. Requires the OAuth scope(s): projects:read (personal access tokens are not scope-restricted).listProjects
ChatGPTname field of each project as the {project} path parameter in project-scoped endpoints like createBuildForProject and listBuilds. Requires the OAuth scope(s): projects:read (personal access tokens are not scope-restricted).listReviews
ChatGPTcreateReview to verify a review was recorded, or to check if a build has already been reviewed before submitting a duplicate. Requires the OAuth scope(s): projects:read (personal access tokens are not scope-restricted).removeBuildCommentReaction
ChatGPTresolveBuildCommentThread
ChatGPTsubscribeBuildCommentThread
ChatGPTunignoreChange
ChatGPTunresolveBuildCommentThread
ChatGPTunsubscribeBuildCommentThread
ChatGPTupdateBuildComment
ChatGPTexecute_typescript
Claudesearch_tools
ClaudeLinks
Similar Apps
Listed on MCP App Store
Add this badge to your site or README. It links to this page.
HTML
<a href="https://mcp-app-store.keenethics-labs.com/app/snapvisor"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>
Markdown
[](https://mcp-app-store.keenethics-labs.com/app/snapvisor)






