Trello MCP With Claude Code: Setup and Everyday Use

To use Trello from Claude Code, add the official Trello MCP server with claude mcp add --transport http trello https://mcp.trello.com/v1, then type /mcp inside Claude Code and sign in to Trello in your browser, choosing the one Trello workspace you want to connect. After that, Claude Code can read your boards, lists and cards, search them, create and move cards, tick off checklists and archive what is finished, all with your own Trello permissions. Atlassian hosts the server and it works on every Trello plan. The rest of this guide covers each step, the prompts that earn their keep, and what the server does not do yet.

The Trello server is separate from the Atlassian Rovo MCP Server that covers Jira and Confluence. If you use both, you add them as two servers; the Jira side is covered in Jira MCP with Claude Code. If MCP itself is new to you, start with what MCP is.

Before you start

• A Trello account on any plan. Atlassian’s documentation↗ (opens in a new tab) says no paid plan is needed; the one exception is creating focus-time events in Planner, which needs Trello Premium or Enterprise.
• Claude Code installed and signed in.
• A browser for the one-time sign-in. On a remote machine, see the section on running without a browser below.
• Your organisation allowing it. On Trello Enterprise, admins control Trello MCP from Atlassian Administration, under Rovo MCP server. They can block the server’s domain or turn off read or write access. If sign-in says access is restricted, that is where the answer is.

Step 1: add the server

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

https://mcp.trello.com/v1 is the URL Atlassian publishes↗ (opens in a new tab) for adding Trello by hand to any MCP client. trello is only the local name, but it becomes part of every tool’s name in Claude Code’s permission rules, so choose it once and keep it. The transport is streamable HTTP, which is what --transport http selects.

Without a --scope flag the server is added in local scope: you only, this project only. Add --scope user to have Trello in every project on your machine, or --scope project to write it to a .mcp.json file that the team commits. With project scope each person still signs in with their own Trello account.

There is a second route. Trello is also listed in Claude’s connector directory. If you add it there and you sign in to Claude Code with the same claude.ai account, Claude Code picks the connector up automatically↗ (opens in a new tab) and it appears in /mcp marked as coming from claude.ai. Use one route or the other, not both.

Step 2: sign in and choose a workspace

1. Start Claude Code and type /mcp.
2. Choose trello. It shows as needing authentication; choose to authenticate.
3. Your browser opens on Atlassian’s sign-in. Sign in with the account you use for Trello.
4. On the consent screen, pick the Trello workspace to connect and review the permissions: read, write and search for boards and cards, plus separate ones for Inbox and Planner. Untick anything you do not want yet.
5. Back in Claude Code, /mcp should show the server connected with its tools listed.

You do not need to create an app or copy a client ID. Trello’s sign-in service publishes the standard OAuth metadata, and Claude Code registers itself from that. What happens during the browser step, and what to look for on the consent screen, is covered in how MCP sign-in works. If you later need a permission you left unticked, Atlassian’s advice is to disconnect and connect again with the wider set.

What it can do today

The tools are grouped by object and switched by an action parameter, so there are few tool names but many operations. The names you will see are trelloReadMember, trelloReadBoard, trelloReadList, trelloReadCard, trelloReadChecklist, trelloReadInbox, trelloReadPlanner, trelloSearch, and the matching trelloWrite… tools. In plain terms:

• Boards: view and create boards, view labels.
• Lists: view and move lists.
• Cards: view, create, update, move, archive and mark done; attach and detach labels.
• Checklists: view, create and update checklists and their items.
• Search: find cards and boards by keyword.
• Inbox and Planner: manage Inbox cards, read calendar events, link cards to events and create focus time.

Atlassian’s repository also ships an optional agent skill↗ (opens in a new tab), trello-use, that teaches an assistant how to call these tools correctly: it must get an object’s id from a read tool before writing, and it converts local times to UTC for due dates. If Claude Code keeps tripping over ids or due dates, installing that skill is the documented fix.

Everyday prompts

Start with reads so you can see how it understands your boards:

• “Who am I signed in to Trello as, and which workspace is connected?”
• “What Trello cards are due this week for me?”
• “Summarise the Sprint board: how many cards in each list, and which are overdue?”

Then the writes that make it worth having in a coding session: “Create a card in To Do on the Sprint board called ‘Retry job double-charges on timeout’ with these steps in the description”, “Move ‘Fix login bug’ to In Progress”, or, at the end of a piece of work, “Add a checklist called ‘Release steps’ to the Release 2.4 card and mark ‘Update changelog’ done.” Ask it to report the card it changed each time, so you can check it in Trello.

Permissions: whose rights does it use?

Yours. Atlassian says every action respects your existing Trello permissions and the assistant cannot do anything you could not do in Trello directly, limited further to the one workspace and the permissions you approved. The server has no permanent delete at all: it can archive cards and lists, which you can restore, but it cannot delete them.

Claude Code adds its own layer. In its default mode it asks before each MCP tool call, and you can pre-approve tools in settings with rules↗ (opens in a new tab) of the form mcp__<server>__<tool>. Allow rules may end in a wildcard after the server name, so one rule can cover every read tool while writes keep asking:

{
  "permissions": {
    "allow": [
      "mcp__trello__trelloRead*",
      "mcp__trello__trelloSearch"
    ]
  }
}

Running it without a browser

Trello MCP signs in with OAuth only. Atlassian says it does not support API token authentication for MCP connections, so there is no header you can pass instead. On a machine without a display, run claude mcp login trello: Claude Code notices there is no browser, prints the sign-in address, and asks you to paste back the redirect URL after you approve on your own machine. Add --no-browser to force that behaviour, and connect with ssh -t so the paste step has a terminal.

Limits to know

• One workspace per connection. Atlassian says multi-workspace support is planned.
• No comments, attachments, custom fields or card activity history yet. All of these are on Atlassian’s published list of coming capabilities, so check its repository before assuming they are still missing.
• Card previews are capped. When a board or card response is marked as having more, the assistant has to page through the full list; ask for counts explicitly rather than trusting a first answer.
• Large results fill context. Claude Code warns when one tool result passes 10,000 tokens and caps it at 25,000 by default (raise it with MAX_MCP_OUTPUT_TOKENS). Narrow the request to one board or list instead.
• It acts as you. There is no separate identity for the assistant, so in Trello its changes are made with your account.

Removing or resetting it

claude mcp logout trello   # clear this machine's sign-in
claude mcp remove trello   # remove the server entirely

The “Clear authentication” option under /mcp does the same as logging out. If you connected through Claude’s directory instead, disconnect it under Customize, then Connectors, in Claude. Atlassian notes you can revoke an assistant’s access at any time, and Enterprise admins can cut it off for everyone from Atlassian Administration.

If Trello is more than you need

Trello gives you any number of lists and a long history of power-ups, and its MCP server works through your own account. If what you want instead is four fixed lanes, a board your assistant joins as a member with a role you chose rather than all of your rights, and a history that names it on every change, that is what fenbs is for. The fair comparison is on fenbs vs Trello, and the Claude Code setup is on its integration page.