TestivAI — Claude Code MCP server: what it does, how to install
TestivAI gives coding agents structured visual-regression evidence, including verdicts and diff images. Use it when UI changes need review beyond the DOM.
claude mcp add mcp -- npx -y @testivai/mcpWhat TestivAI does
TestivAI reads visual test output from visual-report/results.json and exposes it to coding agents through MCP. It can summarize snapshot verdicts, group related changes, show the report location, and return baseline, current, and diff images. Snapshot explanations include pixel regions, DOM and computed-style signals, capture environment, and asset-loading health.
The server provides tools such as get_visual_results, explain_snapshot, get_report, get_diff, list_baselines, approve_snapshot, and approve_all. It also includes the review-visual-changes and triage-visual-diff prompts. Setup uses npx, and no API key is required. A person must confirm an approval before a capture becomes a committed baseline, because an agent approving its own change could preserve a regression.
What it helps with
- review UI changes after a code edit
- separate rendering drift from a real visual change
- inspect a snapshot with readable diff crops
- trace a failed stylesheet or web-font capture
- promote a reviewed capture to the baseline
Example prompts for the agent
- Run
review-visual-changeson the visual test results and give me a recommendation for each snapshot. - Use
triage-visual-diffto classify every diff and missing baseline, then state the next action. - Show the images and explain the snapshot where the DOM stayed the same but the visual result changed.
How to install TestivAI
Platform CLI
claude mcp add mcp -- npx -y @testivai/mcpPlatform CLI
codex mcp add mcp -- npx -y @testivai/mcpConfig file
{
"mcpServers": {
"mcp": {
"command": "npx",
"args": [
"-y",
"@testivai/mcp"
]
}
}
}Config file
{
"servers": {
"mcp": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@testivai/mcp"
]
}
}
}Platform CLI
gemini mcp add mcp npx -y @testivai/mcpStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about TestivAI
How do I install TestivAI in Claude Code?
Run claude mcp add mcp -- npx -y @testivai/mcp. The MCP server will then be available to the project.
What does TestivAI do?
It gives coding agents evidence from visual-regression runs, including snapshot verdicts, diff images, baseline data, and signals from the DOM, styles, environment, and capture health.
Does TestivAI approve visual baselines automatically?
Approval is intended to follow human review. In Claude Code, approve_snapshot and approve_all request confirmation on every approval call.
