Docs

Connect

Ask IA4AI from your AI tool

Add IA4AI to Claude, ChatGPT, Claude Code, Cursor, VS Code, Cline, Windsurf, Gemini CLI or Codex as an MCP connector, then ask it what your AI is doing.

IA4AI is a remote MCP server (Streamable HTTP). Once it is added, your assistant can answer questions like these from your workspace's own records:

  • What is my AI doing? Ask IA4AI.
  • Which of my agents costs the most, and on which models?
  • Is anything wrong with my IA4AI connections?
  • Which of my runs could have used a cheaper model?

The connector reads your workspace's reports. The one thing it can write is a run for investigation (ingest_trace), and it changes no setting. IA4AI sees the tool calls your assistant makes, not your conversation with it. The tools and what each returns are in MCP tools.

#What every tool needs

Two things, in each tool's own format: the server address and your workspace key.

Address
https://ai.odux.io/app/mcp/

Tools that let you set a header send the key as x-api-key: YOUR_IA_KEY. Tools that take only an address (the Claude apps, ChatGPT, Cline's form) get the key in the address instead:

Server URL
https://ai.odux.io/app/mcp/?x-api-key=YOUR_IA_KEY

Warning. An address with the key in it is a credential. Anyone who has it can read your workspace's reports. Don't paste it into a shared document or a screenshot. If it leaks, replace your key under Setup.

Your Connect page shows every block below with your real key in it, and a Check my connection button that turns green once IA4AI has received a call from your tool. Keep the trailing slash on /mcp/: without it, some tools fail to follow the redirect and show no tools.

#Claude

claude.ai and the Claude desktop app

  1. In Claude, open Customize → Connectors (open it). The desktop app has the same menu.
  2. Click Add custom connector. On a Team or Enterprise plan, an Owner adds it for everyone instead, under Organization settings → Connectors → Add → Custom.
  3. Name it IA4AI, paste the server URL below, and leave the OAuth fields empty.
  4. Click Add. Then, in a chat, click + → Connectors and make sure IA4AI is on.
Server URL
https://ai.odux.io/app/mcp/?x-api-key=YOUR_IA_KEY

Check it worked: Ask Claude: “What is my AI doing? Ask IA4AI.”

Note. This URL contains your key. Treat it like a password.

Note. The Free plan allows one custom connector.

#ChatGPT

chatgpt.com, developer mode

  1. In ChatGPT on the web, open Settings → Apps → Advanced settings and turn on Developer mode.
  2. Click Create app.
  3. Name it IA4AI, paste the server URL below, and choose No authentication.
  4. Create it, then turn IA4AI on in a chat.
Server URL
https://ai.odux.io/app/mcp/?x-api-key=YOUR_IA_KEY

Check it worked: Ask ChatGPT: “Use IA4AI: what is my AI doing?”

Note. Developer mode is available on Plus, Pro, Business, Enterprise and Education plans, on the web.

Note. This URL contains your key. Treat it like a password.

#Claude Code

in your terminal

  1. Run this once in a terminal. It adds IA4AI for all your projects.
Terminal
claude mcp add --transport http --scope user ia4ai https://ai.odux.io/app/mcp/ --header "x-api-key: YOUR_IA_KEY"

Check it worked: Run claude mcp get ia4ai, or type /mcp inside Claude Code.

#Cursor

one click

The Add to Cursor button is on your Connect page, with your key in it.

  1. Click Add to Cursor and confirm in Cursor.
  2. Or paste the block below into ~/.cursor/mcp.json. If the file already lists other servers, add only the ia4ai entry inside the existing list: don't replace the whole file.
~/.cursor/mcp.json
{
  "mcpServers": {
    "ia4ai": {
      "url": "https://ai.odux.io/app/mcp/",
      "headers": {
        "x-api-key": "YOUR_IA_KEY"
      }
    }
  }
}

Check it worked: Cursor's MCP settings list ia4ai with its tools.

#VS Code

GitHub Copilot agent mode

  1. Run this in a terminal (macOS or Linux).
  2. On Windows, or if you prefer: run MCP: Open User Configuration from the Command Palette and paste the second block. If the file already lists other servers, add only the ia4ai entry inside the existing list: don't replace the whole file.
Terminal
code --add-mcp '{"name":"ia4ai","type":"http","url":"https://ai.odux.io/app/mcp/","headers":{"x-api-key":"YOUR_IA_KEY"}}'
mcp.json
{
  "servers": {
    "ia4ai": {
      "type": "http",
      "url": "https://ai.odux.io/app/mcp/",
      "headers": {
        "x-api-key": "YOUR_IA_KEY"
      }
    }
  }
}

Check it worked: Run MCP: List Servers: ia4ai is listed as running.

#Cline

in VS Code

  1. In the Cline panel, click the MCP Servers icon, then the Remote Servers tab.
  2. Server name: ia4ai. Server URL: paste the URL below. Transport: Streamable HTTP. Click Add Server.
  3. Prefer the JSON? Configure → Configure MCP Servers and paste the second block. If the file already lists other servers, add only the ia4ai entry inside the existing list: don't replace the whole file.
Server URL
https://ai.odux.io/app/mcp/?x-api-key=YOUR_IA_KEY
cline_mcp_settings.json
{
  "mcpServers": {
    "ia4ai": {
      "type": "streamableHttp",
      "url": "https://ai.odux.io/app/mcp/",
      "headers": {
        "x-api-key": "YOUR_IA_KEY"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Check it worked: ia4ai shows a green dot in Cline's MCP Servers list.

#Windsurf

Cascade

  1. In Cascade, open MCPs → View raw config (the file ~/.codeium/windsurf/mcp_config.json).
  2. Paste the block below. If the file already lists other servers, add only the ia4ai entry inside the existing list: don't replace the whole file.
  3. Save, then click Refresh in the MCP panel.
mcp_config.json
{
  "mcpServers": {
    "ia4ai": {
      "serverUrl": "https://ai.odux.io/app/mcp/",
      "headers": {
        "x-api-key": "YOUR_IA_KEY"
      }
    }
  }
}

Check it worked: The MCP panel lists ia4ai with its tools.

#Gemini CLI

in your terminal

  1. Add the block below to ~/.gemini/settings.json. If the file already lists other servers, add only the ia4ai entry inside the existing list: don't replace the whole file.
settings.json
{
  "mcpServers": {
    "ia4ai": {
      "httpUrl": "https://ai.odux.io/app/mcp/",
      "headers": {
        "x-api-key": "YOUR_IA_KEY"
      }
    }
  }
}

Check it worked: Type /mcp in Gemini CLI: ia4ai is listed.

#Codex

OpenAI's coding agent

  1. Add these lines to ~/.codex/config.toml.
config.toml
[mcp_servers.ia4ai]
url = "https://ai.odux.io/app/mcp/"
http_headers = { "x-api-key" = "YOUR_IA_KEY" }

Check it worked: Run codex mcp list: ia4ai is listed.

#Another tool

any MCP client

  1. Add a remote MCP server (Streamable HTTP) with the address and header below.
  2. If the tool only takes an address, use the address with the key in it instead.
Address
https://ai.odux.io/app/mcp/
Header
x-api-key: YOUR_IA_KEY
Address with the key in it
https://ai.odux.io/app/mcp/?x-api-key=YOUR_IA_KEY

Check it worked: Ask your tool: “What is my AI doing? Ask IA4AI.”

Each tool's format was last checked against its own documentation on 2026-10-22.

#If it does not connect

  • The tool shows no IA4AI tools. Check the address ends in /app/mcp/, with the slash. https://ai.odux.io on its own is the website, not the server.
  • "Couldn't register with the sign-in service." Leave the OAuth fields empty. IA4AI has no OAuth sign-in: the key is in the header or the address.
  • Every call is refused. The key was replaced or mistyped. Create a new one under Setup and update the tool.
  • The assistant answers from memory. Name IA4AI in the question ("Ask IA4AI"), and make sure the connector is switched on for the chat.