For AI agents

1. Open Settings → Connectors and choose Add custom connector.
2. Name it GALTON Assets and set the Remote MCP server URL to the endpoint below.
3. Leave Advanced settings (OAuth) empty — the server is public, no auth. Click Add.
4. In any chat, open the "+" / Connectors menu and enable it, then paste the connect-and-remember prompt above.

https://galton-ai-assets.vercel.app/api/mcp

Note · Claude reaches the server from Anthropic's cloud, so a public URL (this one) is required — a localhost server won't work. Free plan = 1 custom connector.

1. Settings → Apps & Connectors → Advanced settings → turn on Developer mode (Plus/Pro/Business/Enterprise).
2. Back in Apps & Connectors, choose Create.
3. Set the MCP Server URL to the endpoint below, Authentication = No authentication, and Save.
4. Start a chat, attach the connector, and paste the connect-and-remember prompt above.

https://galton-ai-assets.vercel.app/api/mcp

Note · Developer mode surfaces the brand tools directly (no search/fetch contract needed). Business/Enterprise may require a workspace admin to enable it first.

1. If the app can't add a remote MCP server (e.g. the consumer Gemini app), don't fight it — feed it the URLs.
2. Paste the connect-and-remember prompt above; it already lists the read URLs.
3. Or paste the contents of /llms.txt (and /design.md, /voice.md) straight into the chat.

https://galton-ai-assets.vercel.app/llms.txt
https://galton-ai-assets.vercel.app/brand.json
https://galton-ai-assets.vercel.app/design.md
https://galton-ai-assets.vercel.app/voice.md
https://galton-ai-assets.vercel.app/linkedin.md

Note · These endpoints are plain text/JSON with open CORS, so an agent that can browse can fetch them directly.

1. Preferred: Settings → Connectors → Add custom connector (same as the web flow) — paste the URL below.
2. Config-file route: Settings → Developer → Edit Config and add the JSON below. The config file launches local (stdio) servers, so a remote URL is wrapped with the mcp-remote bridge.
3. Restart Claude Desktop; confirm the tools appear under the connectors / 🔌 menu.

{
  "mcpServers": {
    "galton-design": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@latest",
        "https://galton-ai-assets.vercel.app/api/mcp"
      ]
    }
  }
}

Note · macOS config: ~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

1. Add the server (project scope commits it to the repo's .mcp.json):
2. Run /mcp inside a session to confirm it's connected, then prompt with the connect-and-remember prompt above.

claude mcp add --transport http galton-design https://galton-ai-assets.vercel.app/api/mcp
# project scope (shareable, writes .mcp.json):
claude mcp add --transport http --scope project galton-design https://galton-ai-assets.vercel.app/api/mcp

1. Create ~/.cursor/mcp.json (global) or .cursor/mcp.json (project) with the JSON below — a bare url, no type needed.
2. Or use Settings → Tools & Integrations → + Add Custom MCP.

{
  "mcpServers": {
    "galton-design": {
      "url": "https://galton-ai-assets.vercel.app/api/mcp"
    }
  }
}

1. Add .vscode/mcp.json with the JSON below — note the top-level key is servers and the remote uses type: "http".
2. Or run the CLI one-liner; then enable it in Copilot Agent mode.

{
  "servers": {
    "galton-design": {
      "type": "http",
      "url": "https://galton-ai-assets.vercel.app/api/mcp"
    }
  }
}

Note · CLI alternative: code --add-mcp '{"name":"galton-design","type":"http","url":"https://galton-ai-assets.vercel.app/api/mcp"}'

1. Add the server (the remote-HTTP key is httpUrl):
2. Or hand-edit ~/.gemini/settings.json (user) / .gemini/settings.json (project) with the JSON below.

{
  "mcpServers": {
    "galton-design": {
      "httpUrl": "https://galton-ai-assets.vercel.app/api/mcp",
      "trust": true
    }
  }
}

Note · Google is moving consumer tiers from Gemini CLI to the Antigravity CLI (config at ~/.gemini/antigravity/mcp_config.json, where the remote key is serverUrl instead of httpUrl). Check which one your licence uses.

1. Edit ~/.codeium/windsurf/mcp_config.json with the JSON below — the remote key here is serverUrl.
2. In Cascade, open the MCP panel → Refresh.

{
  "mcpServers": {
    "galton-design": {
      "serverUrl": "https://galton-ai-assets.vercel.app/api/mcp"
    }
  }
}

1. Cline → MCP Servers → Configure and add the JSON below — the type must be exactly streamableHttp (camelCase).

{
  "mcpServers": {
    "galton-design": {
      "type": "streamableHttp",
      "url": "https://galton-ai-assets.vercel.app/api/mcp",
      "disabled": false
    }
  }
}

1. Add a context server in ~/.config/zed/settings.json. The reliable form bridges the remote URL through mcp-remote:

{
  "context_servers": {
    "galton-design": {
      "source": "custom",
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://galton-ai-assets.vercel.app/api/mcp"
      ]
    }
  }
}

1. Settings → Tools → AI Assistant → Model Context Protocol (MCP) → Add (As JSON) and paste the JSON below (IDE 2026.1+).

{
  "mcpServers": {
    "galton-design": {
      "url": "https://galton-ai-assets.vercel.app/api/mcp"
    }
  }
}

1. Pass the server as a hosted mcp tool. For a public read-only server use require_approval: "never".

from openai import OpenAI
client = OpenAI()
resp = client.responses.create(
    model="gpt-5.1",
    tools=[{
        "type": "mcp",
        "server_label": "galton_design",
        "server_url": "https://galton-ai-assets.vercel.app/api/mcp",
        "require_approval": "never",
    }],
    input="Summarise the GALTON brand voice and list the design tokens.",
)
print(resp.output_text)

Note · Anthropic/Claude API and the OpenAI Agents SDK have equivalent hosted-MCP tools — same server_url, no auth.