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

# Connect ChatGPT, Claude, Claude Code and Codex

> Let your own AI apps search, read, create and update Cogno Tasks as you, over an OAuth-protected MCP server.

Cogno runs an MCP server for your own AI apps. Add its URL to ChatGPT, Claude, Claude Code or Codex, approve the connection on a Cogno page, and the app works with Tasks in one workspace as you.

Open **Settings → MCP** to copy the server URL and disconnect apps. The steps for each app are below.

## What a connected app can do

| Tool | What it does |
| - | - |
| `search_tasks` | Find Tasks by words, status, Project or assignee |
| `get_task` | Read one Task with its description and link |
| `create_task` | Create a Task |
| `update_task` | Change a Task's title, description, status, assignee, Project or dates |
| `list_projects` | List the workspace's Projects |
| `list_workspace_members` | List members and which one is you |

The app acts with your role. A viewer can search and read but not create or update. Your role is checked on every call, so a role change or leaving the workspace takes effect immediately.

> [!WARNING]
>
> A connected app can see every Task in the workspace you choose, and the Tasks it reads (titles, descriptions, people, dates) are sent to that app's provider.

## Connect an app

**Claude**

1. In Claude, open **Customize → Connectors**, choose **+ Add**, then **Add custom connector**.
2. Name it Cogno, paste the URL from **Settings → MCP** and choose **Continue**.
3. Choose **Connect**, then approve on the Cogno page.

On Team and Enterprise plans an owner adds the connector in **Organization settings → Connectors** first.

**ChatGPT**

1. In ChatGPT, open **Settings → Apps → Advanced settings** and turn on **Developer mode**.
2. Choose **Create app**, paste the URL and pick **OAuth**.
3. Approve on the Cogno page.

Creating and updating Tasks needs a Business, Enterprise or Edu plan. On Plus and Pro the app can only read.

**Claude Code**

```bash theme={"system"}
claude mcp add --transport http cogno <URL>
```

Then run `/mcp` in Claude Code, choose `cogno` and sign in.

**Codex**

```bash theme={"system"}
codex mcp add cogno --url <URL>
codex mcp login cogno
```

## Approve the connection

The Cogno page shows the app's name, what it may do, and where your answer is sent. Choose the workspace and **Allow**. If the answer goes to `localhost` or `127.0.0.1`, it is a program on your computer, such as Claude Code or Codex: allow it only if you started the connection yourself. If the page says the app's name is not verified, the app named itself: allow it only if the destination shown is the app you meant to connect.

## Disconnect

In **Settings → MCP**, choose **Disconnect** next to the app. Its access ends immediately. Each approval is its own connection, so the same app on two computers shows twice and each one disconnects on its own. If you are removed from a workspace, your connections there stop working and do not come back if you are added again.

## How it works

The server is at `/mcp/user` on the Cogno API and follows the MCP authorization spec: OAuth 2.1 with PKCE, Protected Resource Metadata, dynamic client registration and Client ID Metadata Documents. Access tokens last one hour; refresh tokens rotate on every use. A refresh token used again within a minute (an app retrying) gets new tokens; used again later, it ends the connection.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.