Vision MCP Tools
Vision provides MCP (Model Context Protocol) tools for AI assistants to capture screenshots and run visual tests.Available Tools
vision_capture
Take a screenshot of a URL.vision_compare
Compare a snapshot to its baseline.vision_test
Run a full visual test suite.vision_approve
Approve changes and update baseline.vision_list_failures
List failed comparisons needing review.Example Conversations
User: “Take a screenshot of our homepage” Assistant uses:vision_capture with url: "https://example.com"
Response: “I’ve captured a screenshot of your homepage. You can view it here: [link]. The capture was taken at 1280x720 using Chromium.”
User: “Run visual tests on staging” Assistant uses:
vision_test with base_url: "https://staging.example.com"
Response: “Visual tests completed. 8 of 10 pages passed. 2 pages have visual differences that need review. You can review them here: [link]”
User: “Approve the homepage changes” Assistant uses:
vision_approve with the comparison ID
Response: “I’ve approved the homepage changes. The new screenshot is now the baseline (version 4).”