Connect AI tools to Flying Donut with MCP

Use the hosted Flying Donut MCP endpoint to let AI assistants and coding tools securely work with your projects.

What you can do with MCP

Model Context Protocol (MCP) connects AI assistants to external tools and live data. With Flying Donut's MCP server, you can ask your assistant to explore projects, review sprint and backlog cards, inspect tasks, create and update cards, move work between a backlog bucket and a sprint, manage assignments, set due dates, check your due work and reminders, take part in card discussions, follow and correct dependencies, and mark cards as blocked or unblocked without leaving your coding tool.

Due dates work across the card tools: the create tools and flying_donut_update_card take dueDate (the start of the due day with your timezone offset), and clearDueDate removes it; flying_donut_list_cards filters by dueState, dueBefore and dueAfter. flying_donut_list_my_due_cards lists your overdue and upcoming cards, and flying_donut_set_card_reminder, flying_donut_clear_card_reminder and flying_donut_list_my_reminders manage your private reminders. Ask “What am I late on?” or “Remind me about #123 on Friday morning”.

Card discussions work as in the web app.flying_donut_list_card_topics andflying_donut_list_topic_comments read a card's topics and comments;flying_donut_add_card_comment comments in the card's Comments tab, or in a topic when you name one;flying_donut_create_card_topic andflying_donut_update_card_topic open and rename topics; andflying_donut_update_card_comment edits a comment you wrote. Comments notify the card's people and anyone you @mention, and cards show how many topics and comments they have. Ask “What did we decide on #42?” or “Comment on #42 that the fix is deployed”.

Attachments are the files and links on a card.flying_donut_list_card_attachments lists them, andflying_donut_get_attachment hands your AI client an image (up to 5 MB) as an image and a text file, such as a log, notes or a CSV, as text (up to 256 KB). PDFs, Office documents and larger files come back as their details, with where to download them; links to Google Drive or Dropbox are given as links and never opened.flying_donut_upload_card_attachment attaches a file of up to 25 MB, such as notes or a report your AI client wrote, and everyone on the project sees it at once. Cards show how many attachments they have. Ask “What does the screenshot on #42 show?” or “Attach these release notes to #42 as release-notes.md”.

Dependencies come from what your team writes: wherever a card's description or comments, or a task's description, mention another card or task.flying_donut_list_card_dependencieslists a card's dependencies, or one task's withtaskId, filtered by relation andopen. Each row is labelled from the point of view of the card or task asked about, so a card that is Blocked by #12 reads Blocks on #12.flying_donut_set_dependency_type corrects a type, or marks a mention as not a dependency, with your edit rights on the card holding the mention, and the card tools returnwaitingOn: how many cards and tasks that are not done the card is blocked by or depends on. Ask “What is #34 waiting on?” or “#34 doesn't really block #12, make it related”.

For example, ask it to summarize cards in your current sprint or turn a feature idea into a backlog card. The assistant acts as you, so access and changes always follow your Flying Donut permissions.

ChatGPT, Codex and Claude setup

ChatGPT, Codex and Claude connect by signing in to Flying Donut, so you don't need to create or paste an API token. Use the setup steps for your client below with the endpoint https://mcp.flyingdonut.io/mcp, then sign in and approve what it can do.

Codex

Add the hosted MCP server with the Codex CLI:

codex mcp add flying-donut --url https://mcp.flyingdonut.io/mcp

Follow the browser sign-in flow. If sign-in does not start, or you need to reconnect, run:

codex mcp login flying-donut --oauth-client-registration cimd
  1. Open the authorization URL printed by Codex if the browser does not open automatically.
  2. Sign in to Flying Donut, select the projects and permissions to grant, and wait for Codex to confirm success.
  3. Start a new Codex conversation and ask Which Flying Donut projects can you see?

Codex shares MCP configuration between its CLI and IDE extension. In Codex desktop, you can also add this endpoint in the MCP server settings. If you installed the Flying Donut plugin, connect its bundled server and run its setup workflow instead of adding a second connection. See OpenAI's MCP setup guide.

Claude

  1. Open Customize → Connectors, click + Add, then Add custom connector.
  2. Name it Flying Donut, enter the endpoint URL above and click Continue.
  3. Under OAuth client, keep Use Claude's published identity (Recommended). Leave Request headers empty.
  4. Click Add and sign in to Flying Donut when Claude asks.

On Team and Enterprise plans, an owner adds the connector once in Organization settings → Connectors. Each member then clicks Connect in Customize → Connectors and signs in with their own Flying Donut account, so everyone works with their own permissions. See Claude's custom connector guide.

ChatGPT

  1. In ChatGPT settings, open Apps and turn on Developer mode in the advanced settings.
  2. Create a new app, name it Flying Donut and enter the endpoint URL above.
  3. Choose OAuth as the authentication and save.
  4. Sign in to Flying Donut when ChatGPT asks.

In Business and Enterprise workspaces, an admin may need to allow custom apps first. ChatGPT moves these settings from time to time; see OpenAI's developer mode guide for the current steps.

When you connect

Flying Donut opens in your browser and asks you to:

  • Sign in, or confirm the account you are already signed in with
  • Check the app's name and whether it is verified; ChatGPT, Codex and Claude show as Verified
  • Choose the projects the app may use; none are selected until you pick them, and it cannot see any other project
  • Review the permissions below and approve
PermissionWhat the app can do
See projects, members and labelsFind projects and read their members and labels
See sprints, cards and tasksRead sprints, backlog buckets, cards and tasks
Create and change workCreate and update cards, tasks, sprints, buckets and labels; move, assign, block and track time
Delete cards, tasks and labelsDelete cards, tasks and project labels

The app acts as you, so it never gets more than your Flying Donut role allows in those projects. If you lose access to a project, the app loses it too. Projects you join later are not added on their own.

Manage connected apps

  • See every app you connected, its projects and permissions, and when it was last used in Account → Connected apps
  • Revoke an app there and it stops working with Flying Donut right away
  • To add projects or permissions, revoke the app and connect it again
  • The app stays connected while you are away; after a long time without use it asks you to sign in again

ChatGPT, Codex and Claude support this sign-in flow. Other clients, such as GitHub Copilot in Visual Studio Code, Cursor and Claude Code, use an API token as described below.

Connect with an API token

Generate a personal token in Flying Donut under Account → API Access. It is shown only once, so copy it right away. Then use these values in any MCP-compatible client:

  • Transport type: HTTP
  • Endpoint URL: https://mcp.flyingdonut.io/mcp
  • Header: Authorization: Bearer <your-flying-donut-v2-api-token>

Visual Studio Code setup

Create or update your workspace-level .vscode/mcp.json file:

{
  "inputs": [
    {
      "id": "flyingDonutToken",
      "type": "promptString",
      "description": "Flying Donut V2 API token",
      "password": true
    }
  ],
  "servers": {
    "flying-donut": {
      "type": "http",
      "url": "https://mcp.flyingdonut.io/mcp",
      "headers": {
        "Authorization": "Bearer ${input:flyingDonutToken}"
      }
    }
  }
}

See the official Visual Studio Code MCP documentation.

Google Antigravity setup

In Antigravity IDE, open the agent panel's menu, select MCP Servers, then Manage MCP Servers and View raw config. Add the following entry to the existing mcpServers object in mcp_config.json:

{
  "mcpServers": {
    "flying-donut": {
      "serverUrl": "https://mcp.flyingdonut.io/mcp",
      "headers": {
        "Authorization": "Bearer <your-flying-donut-v2-api-token>"
      }
    }
  }
}

Replace the placeholder with your Flying Donut V2 API token, save the file, and refresh the server list. Unlike Visual Studio Code, Antigravity uses serverUrl rather than url.

Visual Studio Code-style inputs and password prompts are not available in Antigravity's documented MCP configuration. For this bearer-token setup, enter your token directly in the headers; it is stored as plain text. Keep the token-bearing configuration private and out of source control.

See the official Google Antigravity MCP documentation.

Other MCP clients

Use the same endpoint and bearer header in your preferred client. Configuration file names and schema differ per tool, so follow each client's official documentation:

Security recommendations

  • Prefer dedicated tokens per tool to simplify rotation and revocation
  • Store tokens in each client's secure input or secret store, not plain-text files
  • Do not commit tokens to Git repositories or share them in issue trackers
  • Rotate tokens immediately if you suspect they were exposed

Ship faster with MCP-powered workflows

Connect Flying Donut to your coding assistant and keep planning and execution in sync.