> Index of every Falcon docs page, for assistants: https://falcon.com/docs/llms.txt

# Connect your assistant

Use Falcon from Claude, ChatGPT, Cursor, VS Code or another AI assistant over MCP. It signs in as you and can do only what you can do.

## Before you start

- An administrator turns assistants on in the **MCP** section, and chooses who may connect one.
- Copy your **server address** from **MCP › Connect**. It is your company's address followed by `/mcp`, such as `https://studio.example.com/mcp`.
- Your assistant must be one your company has set up. An administrator can tell you which, and give you a client ID if your assistant asks for one.
- There is no key to paste. The assistant opens your browser, you sign in, and you choose what it may do.

## Claude Code

Run this in a terminal, then type `/mcp` in Claude Code and choose the server to sign in.

```sh
claude mcp add --transport http my-studio https://studio.example.com/mcp
```

## Claude Desktop and claude.ai

- Open **Settings › Connectors** and choose **Add custom connector**.
- Paste your server address. If an administrator gave you a client ID, enter it under **Advanced settings**.
- Sign in when Claude asks.

## ChatGPT

- Open **Settings › Apps & Connectors**. Under **Advanced settings**, turn on **Developer mode**.
- Choose **Create**, give the connector a name, paste your server address and choose **OAuth**.
- Sign in when ChatGPT asks.

## Cursor

Add this to `~/.cursor/mcp.json`, or to `.cursor/mcp.json` in one project, then sign in from **Cursor Settings › MCP**.

```json
{
  "mcpServers": {
    "my-studio": {
      "url": "https://studio.example.com/mcp"
    }
  }
}
```

## VS Code

Add this to `.vscode/mcp.json` in your project, or run **MCP: Add Server** from the command palette and choose HTTP. Start the server from the file and sign in when asked.

```json
{
  "servers": {
    "my-studio": {
      "type": "http",
      "url": "https://studio.example.com/mcp"
    }
  }
}
```

## Codex and other assistants

For Codex, run these two commands. The second signs you in.

```sh
codex mcp add my-studio --url https://studio.example.com/mcp
codex mcp login my-studio
```

Most other assistants take the same settings as Cursor. **MCP › Connect** shows them with your address filled in.

## What your assistant sees

Many assistants load only a few dozen tools, and every tool costs them attention. So an assistant is offered a small set that still reaches everything:

| Tool | What it does |
|---|---|
| `ask_studio` | Ask in plain words. It reads what it needs as you, then answers or suggests changes, each showing what it would change. Nothing changes until you apply a suggestion. |
| `find_operations` | Find what your assistant can do by what it does. |
| `describe_operation` | One operation's description, the access it needs and exactly what it takes. |
| `run_operation` | Run any operation by name, with the same checks as doing it yourself. |
| `builder_list_agents` | List the agents you can see. |
| `builder_get_agent_run` | Check how a run went. |
| `builder_search_docs` | Search these docs. |
| `builder_read_doc_page` | Read one docs page in full. |

For automation that wants exact calls, an administrator can switch on the tools for whole areas, such as Agents or Workflows, for everyone or for one assistant.

## What it may do

- When you sign in, you choose what the assistant may do: **look**, **change settings**, and **run and test**.
- It can never do more than you can. It gets what you allowed and your role allows, whichever is less.
- Risky changes, such as making a version Live, deleting or anything that sends, wait until you approve them in Falcon.
- Everything it does is in the audit log as you, through that assistant, for example “Dana via Claude Desktop”.
- Disconnect it at any time from **Connections › Yours**. Its access ends at once.

## Your Live agents and workflows as tools

> Coming next: Designed and planned, not built yet.

You will be able to offer a Live agent or workflow to connected assistants as a tool of its own, switched on per agent or workflow, with the same spend limits and approvals as any other channel.

## Related

- [MCP and API guide](https://falcon.com/docs/mcp-and-api)
- [Approve what an assistant asks to do](https://falcon.com/docs/mcp-and-api#assistant-approvals)
- [API and MCP reference](https://falcon.com/docs/api)
- [Troubleshooting](https://falcon.com/docs/troubleshooting)