Run UX research directly from Claude.
ValidateThat's MCP server gives Claude direct access to your account. Describe a study in a chat — Claude designs it, creates it, and gives you the share link. Free within your account limits.
Works with Claude Desktop, claude.ai, and Claude Code — same server URL, same token.
Card sort for the new pricing page nav: Plans, Compare, Enterprise, Education, Pricing FAQ, Contact sales, Self-serve signup, Free trial.
Created. Open card sort "Pricing page nav v2" with 8 cards is live. Share link: validatethat.io/study/abc...
Real interaction — one tool call, one share link.
Concrete workflows
What you can actually say to Claude
Three tools in v0.1. Each one maps to a real moment in a research workflow.
“I'm refactoring our SaaS dashboard nav. Cards are Profile, Settings, Billing, Team, API tokens, Logs, Webhooks, Help, Pricing, Status. Run an open card sort to see how users group these.”
Claude does: Claude calls validatethat_create_card_sort. Study is live in your dashboard, share link in the reply.
“What card sorts and tree tests do I have running right now?”
Claude does: validatethat_list_studies returns id, title, response count, share link for everything in your account.
“How's the homepage IA card sort doing? Should I start analyzing or wait for more responses?”
Claude does: validatethat_get_study_summary pulls response count + completion guidance. Claude tells you whether you have enough data.
Setup
Under two minutes, one config block.
Generate an API token
Settings → Integrations on ValidateThat. Shown once at creation, copy it.
Add to Claude's MCP config
Server URL: validatethat.io/mcp. Paste the token as a Bearer header. Same setup for Desktop / Code / web.
Restart Claude, prompt
The validatethat_* tools become available in any chat. Done.
Full setup walkthrough
Platform-specific instructions for Claude Desktop's config file, the Claude Code CLI, and claude.ai's integrations panel are in our setup guide.
Why MCP, not yet another connector
MCP is the open standard Anthropic built so AI assistants can call real services without per-tool plumbing. We use it ourselves daily — and now you can too.
| ValidateThat MCP | Typical no-code connector | |
|---|---|---|
| Setup time | < 2 minutes | Multi-step Zapier flows |
| Auth model | Bearer API token, same as our Figma plugin | Usually OAuth + scoped credentials per tool |
| Where Claude runs | Desktop / Web / Code — same setup | Tool-specific, manual per platform |
| Cost | Free within your account limits — no upsell for MCP | Often per-call or feature-gated |
Tokens stay yours. So does your data.
API tokens are generated with 32 bytes of CSPRNG entropy, shown to you once, and stored server-side only as a SHA-256 hash. The raw token never sits on our servers.
Anthropic does not see, store, or train on your token — MCP servers run in-process on the Claude side; tokens pass through to us over TLS 1.2+, never to a third party.
Revoke any token instantly from Settings → Integrations. Full security policy at validatethat.io/security.
Try it on a real study today.
Free within your account limits — same 3-study free tier as the dashboard. Generate a token, paste into Claude, prompt away.