wais-mcp

Universal MCP server for WAIS (Web Agent Interaction Standard) — 5 generic tools that work with any WAIS-compatible site.

Works with Claude, ChatGPT, Gemini, Cursor, Windsurf, VS Code Copilot, and any MCP-compatible client.

Tools

Tool	Description
wais_discover	Fetch a site's agents.json and show available actions
wais_register	Register at a site using SD-JWT selective disclosure
wais_execute	Execute any action (search, subscribe, purchase, etc.)
wais_confirm	Confirm high-risk actions or complete payments (402 flow)
wais_status	Check account status, plan, and credits

Quick Start

pip install wais-mcp
wais-mcp init

The setup wizard will ask you to pick your provider (Claude, ChatGPT, Cursor, etc.), choose project or global scope, and enter your API key. It writes the correct config file automatically.

Manual Setup by Provider

If you prefer to configure manually, all providers use the same wais-mcp command — only the config file location and format differ.

Claude Desktop

File: claude_desktop_config.json

{
  "mcpServers": {
    "wais": {
      "command": "wais-mcp",
      "env": {
        "PLATFORM_URL": "https://pod.deeger.io",
        "WAIS_API_KEY": "<your-api-key>"
      }
    }
  }
}

Claude Code

File: .mcp.json (project root) or ~/.claude/settings.json (global)

{
  "mcpServers": {
    "wais": {
      "command": "wais-mcp",
      "env": {
        "PLATFORM_URL": "https://pod.deeger.io",
        "WAIS_API_KEY": "<your-api-key>"
      }
    }
  }
}

ChatGPT Desktop

Requires Developer Mode: Settings > Advanced Settings > Developer Mode.

File locations:

• macOS: ~/Library/Application Support/ChatGPT/mcp-server-config.json
• Windows: %APPDATA%\OpenAI\ChatGPT\mcp-server-config.json
• Linux: ~/.config/ChatGPT/mcp-server-config.json

{
  "mcpServers": {
    "wais": {
      "command": "wais-mcp",
      "env": {
        "PLATFORM_URL": "https://pod.deeger.io",
        "WAIS_API_KEY": "<your-api-key>"
      }
    }
  }
}

Restart ChatGPT after saving.

Gemini CLI

File: ~/.gemini/settings.json (global) or .gemini/settings.json (project)

{
  "mcpServers": {
    "wais": {
      "command": "wais-mcp",
      "env": {
        "PLATFORM_URL": "https://pod.deeger.io",
        "WAIS_API_KEY": "<your-api-key>"
      }
    }
  }
}

Cursor

File: .cursor/mcp.json (project) or ~/.cursor/mcp.json (global)

{
  "mcpServers": {
    "wais": {
      "command": "wais-mcp",
      "env": {
        "PLATFORM_URL": "https://pod.deeger.io",
        "WAIS_API_KEY": "<your-api-key>"
      }
    }
  }
}

Windsurf

File: ~/.codeium/windsurf/mcp_config.json

Or open from Windsurf: click MCPs icon in Cascade panel > Configure.

{
  "mcpServers": {
    "wais": {
      "command": "wais-mcp",
      "env": {
        "PLATFORM_URL": "https://pod.deeger.io",
        "WAIS_API_KEY": "<your-api-key>"
      }
    }
  }
}

VS Code (GitHub Copilot)

File: .vscode/mcp.json (project) or via Command Palette: MCP: Open User Configuration (global)

{
  "servers": {
    "wais": {
      "type": "stdio",
      "command": "wais-mcp",
      "env": {
        "PLATFORM_URL": "https://pod.deeger.io",
        "WAIS_API_KEY": "<your-api-key>"
      }
    }
  }
}

Note: VS Code uses servers instead of mcpServers, and requires the type field.

Usage without MCP (Python SDK)

For custom agents, scripts, or any Python code — no MCP client needed:

import asyncio
from wais_mcp import WAISClient

async def main():
    client = WAISClient(api_key="your-api-key")

    # 1. Discover what the site offers
    site = await client.discover("https://serphub.deeger.io")
    print(site.name, site.list_action_ids())

    # 2. Register (shares only required claims via SD-JWT)
    await client.register(site)

    # 3. Execute actions
    result = await client.execute(site, "search", {"query": "python"})
    print(result)

    # 4. Confirm high-risk actions (if 402 returned)
    # result = await client.confirm(site, challenge_id)

    # 5. Check account status
    status = await client.status(site)
    print(status)

asyncio.run(main())

Works with OpenAI SDK, LangChain, CrewAI, or any Python agent framework.

How It Works

1. Discover — Fetches /.well-known/agents.json from a site
2. Register — Shares only required claims via SD-JWT selective disclosure
3. Execute — Handles tokens, DPoP, endpoint resolution, and async polling automatically
4. Confirm — Polls for completion after 402 confirmation challenges
5. Status — Check credits and plan info

All authentication (PoD tokens, DPoP proofs) is handled transparently.

Configuration

Env Var	Default	Description
PLATFORM_URL	https://pod.deeger.io	WAIS Provider URL
WAIS_API_KEY	(keychain fallback on macOS)	API key from provider dashboard

Development

pip install -e ".[dev]"

License

MIT