> ## Documentation Index
> Fetch the complete documentation index at: https://docs.doploy.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Connect an AI agent

> Give Claude Code, Cursor, Claude Desktop or ChatGPT access to your doploy projects in two minutes

# Connect an AI agent

doploy has a hosted [MCP](https://modelcontextprotocol.io) server at `https://mcp.doploy.io/mcp/`.
Any AI client that speaks MCP can watch your builds, read test results and logs, query your Odoo
data, and, if you allow it, deploy. You sign in with your doploy account in the browser. There are
no tokens to copy.

| Client                    | Setup                         |
| ------------------------- | ----------------------------- |
| Claude Code               | one command, below            |
| Codex CLI                 | one command, below            |
| ChatGPT                   | add a connector in Settings   |
| Cursor                    | one JSON entry                |
| Claude Desktop, Claude.ai | add a custom connector        |
| VS Code with Copilot      | MCP: Add Server               |
| Gemini CLI, Windsurf      | one command or one JSON entry |

<Frame caption="One login covers your projects, builds, logs and Odoo data.">
  <img src="https://mintcdn.com/doploy/72CKAeD-J8JE4APL/images/agents/connected-apps.png?fit=max&auto=format&n=72CKAeD-J8JE4APL&q=85&s=270318a3525b26a482c0413fecd93d18" alt="Connected apps page in doploy settings" width="812" height="1012" data-path="images/agents/connected-apps.png" />
</Frame>

## Claude Code

```bash theme={null}
claude mcp add --transport http doploy https://mcp.doploy.io/mcp/
```

Then, inside Claude Code, run `/mcp`, pick **doploy** and choose **Authenticate**. Your browser
opens the consent screen below.

## Codex CLI

```bash theme={null}
codex mcp add doploy --url https://mcp.doploy.io/mcp/
codex
```

Codex detects OAuth on the server and opens your browser the first time a doploy tool is used.

## ChatGPT

1. Turn on Developer mode: Settings → Connectors → Advanced settings → **Developer mode**.
2. Settings → Connectors → **Create**.
3. Name `doploy`, MCP server URL `https://mcp.doploy.io/mcp/`, authentication **OAuth**, then **Create**.

ChatGPT shows the doploy connector in the composer under Developer mode. Custom MCP connectors need
a Plus or Pro plan on the web.

## Cursor

Add the server to `.cursor/mcp.json` (project) or your global MCP config:

```json theme={null}
{
  "mcpServers": {
    "doploy": { "url": "https://mcp.doploy.io/mcp/" }
  }
}
```

Cursor shows **Needs login** next to the server. Click it to sign in.

## Claude Desktop and Claude.ai

Settings → Connectors → **Add custom connector**. Name it `doploy` and use
`https://mcp.doploy.io/mcp/` as the URL. Custom connectors need a Pro, Max, Team or Enterprise plan.

## VS Code with Copilot

1. Command Palette → **MCP: Add Server** → **HTTP**.
2. URL `https://mcp.doploy.io/mcp/`, name `doploy`, choose Global or Workspace.
3. Command Palette → **MCP: List Servers** → doploy → **Start Server**, then **Allow** the sign-in.

Or add it to `.vscode/mcp.json`:

```json theme={null}
{
  "servers": {
    "doploy": { "type": "http", "url": "https://mcp.doploy.io/mcp/" }
  }
}
```

## Gemini CLI

```bash theme={null}
gemini mcp add --transport http doploy https://mcp.doploy.io/mcp/
```

## Windsurf

Add to `mcp_config.json`:

```json theme={null}
{
  "mcpServers": {
    "doploy": { "serverUrl": "https://mcp.doploy.io/mcp/" }
  }
}
```

## Anything else

Any client that supports remote MCP servers with OAuth works. Point it at
`https://mcp.doploy.io/mcp/` and follow its sign-in prompt.

## Terminal agents

Claude Code, Codex and similar tools also work with the [CLI](/agents/cli). `doploy skill install`
teaches Claude Code when to use it.

## The consent screen

Whichever client you use, doploy asks the same three questions once:

<Frame caption="Access, permissions and project scope. All three can be changed later without reconnecting.">
  <img src="https://mintcdn.com/doploy/72CKAeD-J8JE4APL/images/agents/consent.png?fit=max&auto=format&n=72CKAeD-J8JE4APL&q=85&s=e80fd5c588641814d6a61218e9599d1a" alt="doploy consent screen for an AI client" width="812" height="1488" data-path="images/agents/consent.png" />
</Frame>

1. **Access.** *Your doploy account* is the right choice for coding agents. It covers every project
   you belong to and the Odoo data on every connection you can reach. Odoo permissions still apply
   per connection.
2. **Permissions.** *Read only* is the default. *Read + deploy* additionally lets the agent create
   dev branches, rebuild, restart, update modules and take backups. Production actions always ask
   the agent to confirm first.
3. **Projects.** All projects, or only the ones you tick.

Every choice is editable afterwards in **Settings → Connected apps**, and changes apply on the
agent's next request. See [Permissions and safety](/agents/permissions).

## First things to try

* "Why did the last build on `feature/checkout` fail?"
* "Tail the Odoo log on staging and look for tracebacks."
* "How many open sales orders are in production?"
* "Merge is done. Update the changed modules on staging and tell me when it finishes."

The agent discovers the right tools on its own. The full list is in
[What agents can do](/agents/tools).