Getting Started

Connect an MCP-compatible AI client to Gummble and run your first UI research prompt.

Gummble MCP gives your AI client access to real app screens, flows, and UI patterns. Use it when you want an agent to research production interfaces before it writes copy, designs a flow, or builds UI.

https://mcp.gummble.com/mcp

Requirements

A Gummble account on a plan that includes MCP access
An MCP client with Streamable HTTP support, or a bridge for local-only clients
Browser access for the OAuth consent step

If the client asks for a transport, choose Streamable HTTP. Use https://mcp.gummble.com/sse only for legacy clients that explicitly support SSE but not Streamable HTTP.

Pick your client

Claude Code

Add Gummble with one CLI command.

Cursor

One-click install, or one entry in mcp.json.

Codex

Use Gummble in Codex CLI and the Codex IDE extension.

Claude Desktop

Add Gummble as a custom connector.

Windsurf

Configure Streamable HTTP in Windsurf settings.

v0

Feed production UI references into v0 prompts.

Install flow

Add a server named gummble in your client.
Use https://mcp.gummble.com/mcp as the server URL.
Start the client auth flow.
Sign in to Gummble in the browser window.
Approve the MCP consent screen.
Return to your client and run a prompt that needs UI references.

First prompt

Use Gummble MCP to find 5 mobile onboarding screens from language
learning apps. Compare the first-step layout, CTA copy, and progress
framing.

The client should call Gummble tools and return app names, screen or flow references, and a short synthesis you can use in product work.

Good follow-up prompts

Find examples of annual plan discount framing in mobile paywalls.

Compare search empty states from marketplace apps and summarize the
common recovery actions.

Pull 6 checkout screens with Apple Pay and identify where trust badges
or delivery promises appear.

When something breaks

Most failures are plan, OAuth, transport, or token-scope issues. See Troubleshooting.