How Do You Connect Claude Code to Trello?
Connect Claude Code to Trello by registering Trello's own hosted MCP server, at https://mcp.trello.com/v1, then approving the authorization in your browser the first time it runs. There's no API key to copy or token to generate by hand. Once it's connected, Claude Code can search your boards, read a card's full description and checklist, and create or update cards as part of the same conversation you're already having about a build, instead of you tabbing over to trello.com to check what's actually on the board.
Step One: Register the Server
The setup is a single registration step, the same whether you build inside the Claude Code desktop app or the terminal.
- Run claude mcp add trello --transport http https://mcp.trello.com/v1 to register the server.
- Restart Claude Code. On first use it opens your browser to a Trello authorization screen instead of asking you for a token.
- Approve the connection and pick which permissions to grant, then run claude mcp list to confirm trello shows up as connected.
Read, Write, Search: Pick Scopes on Purpose
Trello's authorization screen breaks permissions into three separate grants instead of one all-or-nothing toggle, and that separation is the whole safety model. Treat it the same way you'd treat a database credential: hand over exactly what the task needs, not the full set out of convenience.
What each Trello MCP permission actually unlocks
| Permission | What it lets Claude Code do | Grant it when |
|---|---|---|
| Read | View existing boards, lists, cards, and checklists as context | Almost always - this is the baseline for pulling a spec out of a board |
| Search | Find relevant cards across boards without you naming the exact one | You have more than a handful of boards and don't want to paste links every time |
| Write | Create and update cards, lists, and checklists using your Trello permissions | You actually want Claude Code moving cards and leaving updates, not just reading |
Start with Read and Search only. Add Write later, deliberately, once you've watched the connection behave for a week and trust what it does with a card it's allowed to touch.
What This Is Actually Good For
The obvious use is pulling a real spec instead of retyping one. Point Claude Code at a card and it reads the description and checklist directly, the same pattern as [connecting Claude Code to Notion](/blog/connect-claude-code-to-notion) or Linear, just for boards instead of docs or tickets.
- Pull a card's full description and checklist as the spec for a build, instead of summarizing it into a prompt yourself and losing the detail buried in a comment.
- Have Claude Code move a card to Done and leave a comment with what shipped once a change is finished, so the board reflects reality without you switching tabs to update it.
- Turn a running board into a plain-language status update a client can read, useful when they check Trello and you'd otherwise be writing that summary by hand every Friday.
None of this replaces actually looking at the board yourself before a client call. It removes the manual copy-paste tax between where the task list already lives and the session where you're doing the work.
The One Thing It Still Won't Do
That's a meaningfully different safety model than the read-only database bridge covered in [how to connect Claude Code to a database](/blog/connect-claude-code-to-a-database), where the protection lives in the credential you create. Here the protection is built into the server itself. It's a reasonable default for a task board, where the worst outcome of a bad prompt is clutter you can clean up, not data you can't get back.
Skip the Older Community Servers Now
When Not to Bother With It
Skip the setup if your task list already lives inside Claude Code itself, or if it's a solo project with no board to speak of. It earns its place the moment a real board is the source of truth for a project, and you'd otherwise be tabbing over to trello.com every few minutes to check what's actually assigned and what's actually done.
Short, practical drops on skills, MCP, agents, prompts, and more. No spam, unsubscribe anytime.
Frequently asked questions
How do I connect Claude Code to Trello?
Run claude mcp add trello --transport http https://mcp.trello.com/v1, restart Claude Code, and approve the browser authorization screen the first time it runs. No API key or token generation is required with the official route.
Do I need a Trello API key and token to connect it?
No, not with the official hosted server. Trello's own MCP server at mcp.trello.com/v1 uses a browser authorization flow instead. A Trello API key and token are only needed for older, community-built Trello MCP servers.
Can Claude Code delete cards or boards once it's connected?
No. Trello's MCP server has no delete tool at all, regardless of which permissions you grant. A write-enabled connection can create, move, and update cards, but it cannot permanently remove a board, list, or card.
Which permissions should I grant when connecting Claude Code to Trello?
Grant Read and Search by default, since those cover pulling context out of boards. Only add Write once you specifically want Claude Code creating or updating cards, not just reading what's already there.
I already set up a community Trello MCP server. Should I switch?
It's worth switching to the official mcp.trello.com/v1 server when you get a chance. It's maintained directly against Trello's own product, and the browser authorization is easier to scope and revoke than a long-lived API token sitting in a config file.
Last reviewed by David Iya on September 2, 2026
