Does the AI run the tests itself?

The assistant writes and manages tests; Validate.QA executes, heals, and wires them into CI. That split keeps execution consistent no matter which IDE you use.

Use it from your IDE (MCP)

Q: Which IDEs and assistants are supported?

Any MCP-capable assistant. Chat connectors (Claude Desktop, claude.ai, Grok, Grok Bot, Codex) use the hosted /mcp URL with OAuth. AI IDEs (Claude Code, Cursor, VS Code, Codex CLI) use the stdio package with a pat_ token.

Chat connectors (Claude, Grok, Grok Bot, Codex) use the hosted /mcp URL with OAuth. AI IDEs use npx @validate.qa/mcp with a pat_ token. Same 48 tools either way.

If you live in an AI IDE or chat app, you don't have to leave it. Validate.QA exposes a Model Context Protocol (MCP) surface so Claude, Grok, Grok Bot, Codex, Claude Code, Cursor, and VS Code can drive the whole platform — the AI writes and manages tests, and Validate.QA handles execution, healing, and CI.

Configure it (AI IDEs)

Add the stdio server to your IDE's MCP config with your project's API URL, a personal access token (pat_…), and the project id. The package runs over stdio via npx — there's nothing to install globally. Runner tokens (urt_ / rt_) are not accepted for MCP.

MCP server config (e.g. in your IDE's mcp.json)

Connect Claude, Grok, Grok Bot, or Codex

Chat apps and connector catalogs use the hosted URL — no PAT and no npx. In Project Settings → MCP, copy {your API}/mcp. Claude Desktop / claude.ai: Customize → Connectors → Add custom connector. Grok: grok.com/connectors → Custom. Grok Bot: Settings → Plugins. Codex: add a custom MCP connector with the same URL. Sign in, pick one project on the consent screen, and the assistant gets the same 48 tools plus live project resources over OAuth. Shared tenants use https://shared.validate.qa/mcp; dedicated orgs copy their own host. Walkthrough: /blog/connect-claude-grok-codex-to-validate-qa.

What the AI can do

Create, edit, list, run, and heal tests — without leaving your editor.

Set up a whole project (features, URL, credentials) in one call.

Run a security audit and list findings and discovered API endpoints.

Create schedules and read live project state — pass rates, runner health, and coverage gaps.

Live project context

The server also exposes live resources the assistant reads first — a project overview (test counts, pass rates, runner health, feature gaps) and your test conventions (credential patterns, tag meanings) — so generated tests match how your project already works.

Open Use it from your IDE (MCP) · Get Started Free