How do I connect Claude Code to Trello?

Use Trello's official MCP server. In your project, run claude mcp add --transport http trello https://mcp.trello.com/v1, then open /mcp in Claude Code and sign in through Trello's OAuth consent screen. If you log in to Claude Code with a claude.ai account, you can also add the Trello connector on claude.ai, and Claude Code picks it up.

Atlassian announced the official Trello MCP server on July 22, 2026. Everything below reflects its documentation as of October 23, 2026.

What the Trello MCP server is

MCP (Model Context Protocol) is how Claude Code connects to outside tools. Trello hosts its own server at https://mcp.trello.com/v1. It is separate from Atlassian's Rovo MCP server, which covers Jira, Confluence and other Atlassian products but not Trello.

According to Atlassian's docs:

• Authentication is OAuth only. You approve access on a consent screen. Trello does not currently support API tokens for MCP.
• Access is scoped to the workspaces you pick. You can select one or several on the consent screen.
• Your Trello permissions apply. The assistant can't do anything you can't do in Trello.
• No permanent deletes. It can archive cards and lists, not delete them.
• All Trello plans can use it. Custom fields need Standard, Premium or Enterprise.

Step 1: add the server to Claude Code

Trello's FAQ says to add the server URL to your client's MCP settings. In Claude Code, a remote server is added with --transport http:

claude mcp add --transport http trello https://mcp.trello.com/v1

This uses the default local scope: the server loads only in the current project and stays out of version control. To use Trello in every project:

claude mcp add --transport http trello --scope user https://mcp.trello.com/v1

To share the config with your team through the repo, use --scope project, which writes .mcp.json at the project root.

Step 2: authorize it

Start Claude Code and run:

/mcp

Select the Trello server and follow the sign-in in your browser. On Trello's consent screen, choose the workspaces and the permissions (Read, Write, Search) to grant. Claude Code stores the tokens and refreshes them automatically. You can also run the flow from your shell:

claude mcp login trello

If you later need a permission you didn't grant, Atlassian's advice is to disconnect Trello MCP and reconnect with the new permissions. In Claude Code, "Clear authentication" in the /mcp menu revokes the stored access.

Alternative: the claude.ai connector

If you log in to Claude Code with a claude.ai account, connectors you add on claude.ai are available in Claude Code automatically. Trello is listed in the Claude connector directory. Add it at claude.ai, authorize it there, and it appears in /mcp with a marker showing it comes from claude.ai. On Team and Enterprise plans, only admins can add connectors.

Step 3: use it

Ask in plain language:

List the cards in the "To do" list of the Backend board.

Read the card "Fix CSV export", implement it on a new branch, then move the card to "Review" and add a checklist of what you changed.

Atlassian's current tool list covers boards (view, create), lists (view, move), cards (view, create, update, move, archive, mark done, labels, members), checklists, Inbox, search, activity history, attachments (read, attach links) and custom fields. Attachments upload, label creation and copying cards are listed as coming later.

Limits of this method

• Claude Code doesn't know which card belongs to which session. With several agents open, you track that mapping yourself, or ask each agent to comment on its card.
• One more auth surface. Each connection is an OAuth grant per workspace, managed in /mcp or on claude.ai.
• Archive, not delete. Cleanup of test cards is manual.
• The card is untrusted input. Whatever is written on a card enters the agent's prompt. Review cards written by others before an agent acts on them.
• Enterprise controls. On Trello Enterprise, admins can restrict or turn off AI access from the Rovo MCP section of Atlassian Administration. If the connection fails at work, check that first.

With Kanbell

Kanbell does not integrate with Trello. It can't read or move Trello cards, and it doesn't ship a Trello MCP server. If your tasks live in Trello, the method above is the way to go.

What Kanbell has instead is its own task board per project, either local or backed by GitHub Issues (columns open/closed or by labels). From a card, you launch an agent in a real terminal: the brief is pasted, not sent, and you press Enter. Agents can read the board through the kanbell CLI (kanbell list, kanbell show). It's an alternative if you are willing to keep tasks next to the code, in the app or in GitHub, rather than in Trello. See how to connect Claude Code to GitHub Issues and how to organize tasks for AI coding agents.