How to Add the Trello MCP Server to Claude Desktop

Setting up the Trello MCP server in Claude Desktop allows you to manage your boards and cards directly through the AI interface. To begin, you will need to ensure your environment is configured correctly, which often involves installing Node.js and obtaining your Trello API credentials.

While some users prefer to automate these workflows using Activepieces to connect various apps, the manual setup for Claude requires editing your configuration file to include the Trello server details.

Once the JSON configuration is updated with your API key and token, Claude will be able to execute commands like creating cards or listing board members seamlessly.

The Trello MCP server is a standardized interface that enables Claude Desktop to interact directly with Trello boards, lists, and cards by integrating API credentials into the Model Context Protocol configuration.

What the Trello MCP server enables

By creating a persistent, bidirectional link between your local Claude Desktop environment and your Trello workspaces, the Trello Model Context Protocol (MCP) server changes how you interact with your tasks. This connection allows Claude Sonnet 5.5 to manipulate project data without manual copy-pasting.

Exposing Trello’s API endpoints directly to the model’s reasoning engine allows you to move beyond simple chat interactions into an agentic workflow. The AI can autonomously organize boards, update due dates, and sync tasks.

Requirements for the Trello MCP server setup

Success with this integration depends on a specific stack of local runtime tools and cloud credentials that grant Claude the necessary permissions to act on your behalf. Before configuring the local environment, ensure you've gathered the following components.

To execute the MCP server logic on your local machine, you will need Node.js (v18+). You also need a Trello account with at least one active Workspace, providing the foundational data structure for the agent to manage.

The unique identifier for your specific application instance is the Trello Developer API Key. A Trello Member Token allows the server to operate under your specific user permissions.

Finally, you need Claude Desktop for Mac or Windows, the host application that manages the local configuration file and the model interface.

Why MCP replaces traditional browser extensions

Standard browser extensions are limited to DOM-based scraping by the Trello web interface. The MCP server provides Claude with structured access to the underlying API.

While extensions often break when a UI updates, the MCP server relies on stable data schemas. A project lead can rely on Claude Opus 5.5 to generate accurate status reports even if the visual layout of a board changes.

How Trello MCP server permissions stay secure

This architecture also solves the "least privilege" design problem. Instead of giving a third-party extension full access to your browser history, you provide the MCP server with a scoped token.

Only the specific Trello boards you authorize will interact with this token. By shifting the integration from a visual overlay to a protocol-level connection, you enable long-horizon agentic work. This work remains functional and secure across different operating systems.

By shifting the integration from a visual overlay to a protocol-level connection, you enable long-horizon agentic work.

Step 1: Generate your Trello developer credentials

Establishing a secure handshake between Claude Sonnet 5.5 and your workspace requires specific API credentials that bypass the standard user interface.

Moreover, while many teams currently rely on personal API keys shared via insecure internal documentation, the sanctioned method involves creating a dedicated Power-Up to isolate the agent's permissions.

Accessing the Trello Power-Up Admin Portal

The Trello Power-Up Admin Portal is the centralized management hub for workspace integrations. It allows you to define the scope of an agent’s access without exposing your entire account.

Moving away from broad, high-risk tokens toward a managed integration begins by navigating to this portal. By creating a workspace-specific Power-Up here, you ensure that Trello logs the agent's activity and allows you to revoke access instantly without affecting other third-party tools.

Generating the API Key and Manual Token

Securing the connection requires a two-part credential set that defines both the application identity and the specific user authorization.

To generate Trello credentials:

1. Log into the Power-Up Admin Portal and create a new 'MCP' Power-Up.
2. Copy the 32-character API Key, click the 'Token' link to authorize, and save the resulting 64-character token.

This manual token generation gives the Model Context Protocol (MCP) server the explicit right to act on behalf of your user account within the constraints of the Trello API.

Because this token doesn't expire by default, it eliminates the frequent re-authentication that often leads users to abandon secure workflows. Once you secure these strings, you have the components to configure the local environment file that directs Claude’s reasoning capabilities toward your project boards.

Configure the Trello MCP server settings

Integrating the Model Context Protocol (MCP) requires modifying the underlying configuration file that dictates which external tools are available to the model during a session.

While many teams currently rely on manual copy-pasting of project status into a chat window, establishing a formal server connection ensures that Claude Sonnet 5.5 or Claude Opus 5.5 can query live card data without user intervention.

Locating your config.JSON file

The configuration file resides within the application support directory of your operating system. It acts as the bridge between the local Claude client and the external Trello environment.

You will find the file at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS, while Windows users can find it at %APPDATA%\Claude\claude_desktop_config.json.

Accessing this file is the only way to register new capabilities, as the desktop interface doesn't currently offer a graphical menu for server management.

Mapping the Trello API keys to environmental variables

To authorize the connection, you must insert a JSON block into the mcpServers object that defines the runtime environment for the Trello handler. This block requires the absolute path to the Trello server executable and an env object containing your specific credentials.

The primary identifier that associates the request with your developer account is the TRELLO_API_KEY. The TRELLO_TOKEN is the authorization string that grants the agent permission to act on behalf of your user profile.

Formatting the configuration block

The configuration must follow a strict JSON structure to be parsed by the desktop client. You must nest the trello entry inside the mcpServers object, ensuring that the npx command points to the official server package.

{
 "mcpServers": {
 "trello": {
 "command": "npx",
 "args": [
 "-y",
 "@modelcontextprotocol/server-trello"
 ],
 "env": {
 "TRELLO_API_KEY": "your_api_key_here",
 "TRELLO_TOKEN": "your_token_here"
 }
 }
 }
}

The following configuration example demonstrates a successful handshake where the system validates the token.

The status code in the logs indicates that the environmental variables are correctly formatted, which allows the agent to proceed with board discovery.

Restarting the Claude environment

Claude only parses the configuration file during its initial boot sequence. The application ignores any changes made to the JSON file while it is active. You must fully quit the application and relaunch it to initialize the Trello server.

A hammer icon appears in the chat interface once you have restarted. This indicates that the agent now has the authority to browse lists and move cards across your workspace.

Verify the Trello MCP server connection

You verify the Trello MCP connection by looking for the tool icon in the Claude interface. This confirms the Model Context Protocol has successfully handshake with your local configuration.

Confirming the connection and permissions

Claude Sonnet 5.5 has initialized the Trello server if the hammer icon is present in the message bar. It has also mapped the available JSON-RPC methods to its internal reasoning engine.

If the icon is missing, the configuration file likely contains a syntax error, which prevents the environment from mounting the executable.

Clicking this icon allows you to audit the specific permissions granted to the agent. This check helps you avoid exposing private boards to the model's context window by mistake.

This manual check is the primary gatekeeper for maintaining a least-privilege environment before you begin delegating task management.

Syntax for creating and moving Trello cards

Claude manages your workspace by translating natural language intent into structured API calls that target specific board and list IDs. When you provide a prompt, the model selects the appropriate tool from the following set to execute the requested change:

The trello_create_card tool generates a new card in a specified list, allowing the agent to capture action items from a conversation without manual entry.

The trello_get_board_cards tool retrieves all cards from a board, providing the model with the necessary context to summarize project status or identify bottlenecks.

The trello_move_card tool updates the list ID of an existing card. This enables the agent to advance tasks through a "Doing" to "Done" pipeline based on your verbal confirmation.

The trello_search_cards tool queries the workspace for specific strings, which prevents the creation of duplicate tasks by checking existing records first.

Extracting hidden board and list identifiers

Finding the specific IDs required for these tools is not possible through the standard Trello buttons. You can find a Board ID by opening the board in your browser and appending .json to the end of the URL.

The browser will display a text file where the id field at the very top contains the string you need.

To find a List ID, you must look further into that same JSON view or use the Trello API directly.

You can also find these IDs by asking Claude to list your boards once the MCP server is active, as the model can see the metadata that remains hidden from the human interface.

Fixing common Trello MCP connection errors

Troubleshooting Trello MCP errors requires verifying the specific handshake between the local Node.js environment and the Atlassian API gateway.

While users often default to broad permissions, the Model Context Protocol (MCP) fails immediately if the local configuration file contains a single misplaced character or an expired credential.

Fixing Trello MCP authentication errors

A 401 error signifies that the Trello API has rejected the provided credentials. This rejection prevents Claude Sonnet 5.5 from accessing your boards. This typically stems from a mismatch in the claude_desktop_config.json file where the API key or token is incorrectly mapped.

Because the Trello MCP server requires both a developer key and a user-specific token, omitting either one results in a failed authentication loop. You must ensure the "TRELLO_API_KEY" and "TRELLO_TOKEN" environment variables are explicitly defined within the env block of your config.

Trello API rate limits per 10 seconds

Trello enforces strict request caps to maintain platform stability, which can trigger 429 errors during high-volume agentic tasks.

Limit Type	Requests	Time Interval
API key limit	300 requests	10 seconds
API token limit	100 requests	10 seconds

300 requests per 10 seconds is the limit for the API key, according to Atlassian. This means an organization-wide tool could be throttled if multiple users share a single developer key.

The API token limit is 100 requests per 10 seconds, so an individual agent performing deep recursive searches across dozens of cards will hit a ceiling faster than the global key. Exceeding these thresholds forces Claude to pause execution, potentially timing out the reasoning chain.

Reading Claude Desktop logs for Trello MCP errors

If the interface remains unresponsive, the local logs provide the only definitive trace of whether the "npx" command is correctly initiating the server.

A JSON syntax error, such as a missing comma after a tool definition, will prevent the entire MCP plugin from loading. This renders the Trello integration invisible to the model.

By checking the logs, you can distinguish between a path error and a logic error. In a path error, Node.js cannot find the server files. In a logic error, the model lacks the specific board ID needed to execute a command.

Scaling Trello automation beyond a single chat window

Scaling Trello automation requires shifting from local, session-based interactions to a persistent infrastructure. This infrastructure connects Claude Sonnet 5.5 to your broader stack.

While the Model Context Protocol (MCP) provides a direct line for a single user to query a board, it tethers the intelligence to the active chat window.

Activepieces treats agents as actors with least privilege, providing a record of every tool they call.

By placing every agent within the same RBAC and SSO model as employees, companies like MoneyGram and FundingSocieties maintain audit logs for every individual tool call rather than collapsing them into opaque results.

What Activepieces does about this

Activepieces provides a managed environment for the Model Context Protocol that eliminates the need for manual JSON configuration and local Node.js maintenance.

By moving the Trello MCP server from a local config file to a cloud or self-hosted instance, you enable Claude to access your project data across any device, including the mobile app.

This architecture replaces the fragile local handshake with a persistent gateway that remains active even when your desktop application is closed.

The platform handles the authentication mapping through a visual interface, allowing you to connect Trello credentials to the MCP server without exposing plain-text tokens in a local directory.

Under the MIT License, Activepieces allows teams to self-host this infrastructure, ensuring that the bridge between the LLM and the Trello API stays within a private network.

This setup prevents the common 401 and 429 errors by managing request pooling and credential rotation at the server level.

Every connector is an agent tool when managed through Activepieces. Once an integration is registered, it functions simultaneously as a flow step and a tool schema on a per-project MCP server, accessible by Claude, ChatGPT, or custom agents without a second migration or re-integration step.

This unified approach allows a project lead to build a workflow where Claude not only moves a Trello card but also triggers a Slack notification or a GitHub update through the same protocol-level connection.

By using Activepieces as the host for your MCP servers, you gain a centralized dashboard to audit every action the agent performs.

Instead of relying on local logs to debug a failed card creation, you can view the execution trace of the Trello tool call in real-time.

This visibility is why organizations like Contentful use the platform to bridge the gap between generative AI and structured business data, ensuring that agentic workflows are as reliable as traditional automation.

Frequently asked questions

Can I use Trello MCP on the Claude mobile app?

Mobile support for Model Context Protocol (MCP) depends entirely on where the local host environment resides. The Claude mobile app cannot currently initiate a connection to a desktop-bound MCP server.

The mobile interface lacks the filesystem access and process management required to bridge to your Trello data because the protocol relies on a transport layer between the Claude Desktop application and a local process.

Users requiring mobile coordination must rely on the standard Trello app, as the agentic bridge is restricted to the machine where the configuration file and server environment are physically hosted.

Is my Trello API token stored securely?

Security for your Trello credentials is a matter of local environmental control rather than cloud-based storage. The token remains within your machine's configuration files.

When you map your Trello API key and Token to the Claude Desktop configuration, you're placing them in a plain-text JSON file.

Consequently, any user or process with read access to your local profile can intercept these credentials.

To mitigate this risk, you should apply OS-level permissions to the configuration directory. This ensures that only your specific user account can access the secrets required to authenticate the server.

How do I add multiple Trello accounts to one MCP server?

Managing multiple Trello identities requires defining unique server instances within your configuration file. The standard Trello MCP implementation binds a single API token to a single server process.

By creating distinct entries in the mcpServers object, you can name them to reflect the specific account, such as trello-work and trello-personal.

Assign the unique API key and Token pairs to the environment variables of their respective server entries. Restart the Claude Desktop application to initialize the separate processes, allowing Claude Sonnet 5.5 to route queries to the specific board context based on your natural language instructions.

Related reading

• Twilio MCP Server Setup for Claude in 2026
• Connect Claude to Webflow with the MCP Server
• Playwright MCP server for Claude Code and Cursor

References

• Atlassian