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

# Connect your assistant

> Add the Glood MCP server to Claude, ChatGPT, Codex, Gemini CLI, or opencode

One URL, every client:

```
https://platform.glood.ai/api/v1/mcp
```

**There is no key to paste.** The server registers each client dynamically, so
nothing you install is a credential. Your first tool call opens a browser: pick
which of your organizations that connection may reach, approve, and the client
finishes on its own.

<Tip>
  Using a client that is not listed below? Any assistant that speaks the Model
  Context Protocol can connect. See [Connect any MCP client](/platform/mcp-server).
</Tip>

## Pick your client

<Tabs>
  <Tab title="Claude Code">
    Install the plugin. It carries the MCP server **and** the two skills that
    teach Claude when to use it, so there is nothing else to add:

    ```bash theme={null}
    /plugin marketplace add LoopClub/glood-claude-plugin
    /plugin install glood
    ```

    Or add the server on its own, without the skills:

    ```bash theme={null}
    claude mcp add --transport http glood https://platform.glood.ai/api/v1/mcp
    ```
  </Tab>

  <Tab title="claude.ai">
    **Settings → Connectors → Add custom connector**, with the URL above.

    Nothing to install. claude.ai does not load `SKILL.md` files, so the server
    sends the same guidance itself when the connection opens.
  </Tab>

  <Tab title="ChatGPT">
    **Settings → Connectors**, in developer mode, with the URL above.

    Nothing to install, for the same reason as claude.ai: the server sends its
    guidance when the connection opens.

    Building a Custom GPT instead? See [Custom GPT](#custom-gpt-and-actions)
    below — it is the one path that works differently.
  </Tab>

  <Tab title="Codex">
    Add the server to `~/.codex/config.toml` (or `.codex/config.toml` in a
    trusted project):

    ```toml theme={null}
    [mcp_servers.glood]
    url = "https://platform.glood.ai/api/v1/mcp"
    ```

    A `url` instead of a `command` is what selects the HTTP transport. Then
    sign in once:

    ```bash theme={null}
    codex mcp login glood
    ```

    Add the skills too, so Codex knows how to use what it now has:

    ```bash theme={null}
    npx skills add LoopClub/glood-agent-skills
    ```
  </Tab>

  <Tab title="Gemini CLI">
    Add the server to `~/.gemini/settings.json`:

    ```json theme={null}
    {
      "mcpServers": {
        "glood": { "httpUrl": "https://platform.glood.ai/api/v1/mcp" }
      }
    }
    ```

    Then add the skills:

    ```bash theme={null}
    npx skills add LoopClub/glood-agent-skills
    ```

    <Warning>
      The Gemini consumer app has no custom connectors and cannot connect. Gemini
      CLI and Gemini Enterprise can.
    </Warning>
  </Tab>

  <Tab title="opencode">
    Add the server to `~/.config/opencode/opencode.json`:

    ```json theme={null}
    {
      "mcp": {
        "glood": {
          "type": "remote",
          "url": "https://platform.glood.ai/api/v1/mcp",
          "enabled": true
        }
      }
    }
    ```

    Then add the skills:

    ```bash theme={null}
    npx skills add LoopClub/glood-agent-skills
    ```
  </Tab>
</Tabs>

## Approving the connection

The first tool call opens your browser at the Glood consent screen.

<Steps>
  <Step title="Tick the organizations this connection may reach">
    Only what you tick. An organization you leave unticked is invisible to this
    connection, and the assistant has no tool that can add one.
  </Step>

  <Step title="Approve">
    The client finishes on its own — nothing to copy back.
  </Step>

  <Step title="Ask it what it can see">
    > "What Glood organizations can you see?"

    A good answer names each organization, your role in each, and the stores
    available to you. Make this the first question in every new conversation.
  </Step>
</Steps>

## Custom GPT and Actions

Building a Custom GPT rather than using a ChatGPT connector? Import the
OpenAPI description in the GPT builder:

```
https://platform.glood.ai/api/v1/actions/openapi.json
```

<Warning>
  **This is the only path that needs credentials.** The GPT builder cannot
  register itself dynamically, so it needs a client id and secret issued against
  ChatGPT's fixed redirect URI. Email [support@glood.ai](mailto:support@glood.ai)
  for one rather than trying to register it yourself.
</Warning>

## Self-hosted or staging

The URL is the only thing that changes. Point your client at your own API host;
discovery, consent and the token exchange all follow from it.

## Next

<CardGroup cols={2}>
  <Card title="Install the skills" icon="book-open" href="/platform/skills">
    What they teach, and why claude.ai and ChatGPT need nothing installed
  </Card>

  <Card title="Using it" icon="comments" href="/platform/using">
    What to ask for, and how to read the answers
  </Card>
</CardGroup>