How to connect a Trello MCP with Codex (4 steps)

Task cards in Trello hold more context than what developers copy into a Codex task description.

The card description might define acceptance criteria. The checklist might document the implementation steps the team agreed on. The comment thread might contain the edge cases a reviewer flagged, etc.

But Codex writes code against the task description, not the card. This means that when the card has context the developer didn't transcribe, it doesn't shape the output.

The generated fix misses the edge case from the comment thread. And the generated implementation doesn't follow the checklist order the team decided on.

To give Codex direct access to Trello as it works through your coding tasks, we'll show you how to connect Trello with Merge Agent Handler's Trello MCP server.

How it works

Merge Agent Handler connects Codex to the Trello API through the Merge CLI. You install the CLI, authenticate once, and run a single setup command from your project root.

That command writes a Merge CLI section to your project's AGENTS.md file, which tells Codex when to call merge search-tools and merge execute-tool to reach Trello.

Once connected, Merge handles Trello OAuth token storage and refresh on your behalf, so no API keys or credential state live in your repo.

Related: How to use the Trello MCP in Claude Code

Prerequisites

Before getting started, you'll need the following:

• A Merge Agent Handler account
• Codex access (available via the OpenAI platform)
• pipx installed (run pipx --version to confirm, or install via pip install pipx)
• A Trello account with access to the boards you want to connect

If you want to connect Merge Agent Handler's Trello MCP with internal or customer-facing agentic products, you can follow the steps in our docs.

1. Install the Merge CLI

Install the Merge CLI with pipx: pipx install merge-api

Verify your installation: merge --version

2. Log in to Merge

Authenticate the CLI with your Merge Agent Handler account: merge login

This links your CLI session to your Merge account.

3. Add Agent Handler to Codex

From the root of the project where you want Codex to have access to Trello, run:

merge setup agents-md

This writes a Merge CLI section to your project's AGENTS.md file so Codex knows to use the CLI when a task requires Trello data. The command is idempotent, safe to re-run if you need to reset the configuration.

Commit the updated AGENTS.md so the configuration travels with the repo.

Related: A guide to integrating the Trello MCP with Cursor

4. Authenticate Trello

Create a Codex task that requires live Trello data.

For example: "Read the bug card for the payment retry failure, including all comments and checklist items, and generate a fix with a regression test that covers every failure condition documented on the card."

The first time Codex invokes a Trello tool, a Magic Link will appear to complete connector authentication.

Once authenticated, Codex has access to your Trello boards through Merge for all subsequent tasks in this project.

{{this-blog-only-cta}}

Trello MCP FAQ

In case you have more questions on setting up and using the Trello MCP with Codex, we've addressed several more commonly-asked questions below.

What can you do once the Trello MCP is connected to Codex?

With Trello connected, Codex can:

• Read a bug card before generating a fix: pull the card description and comment thread where the issue was scoped and discussed, so generated code addresses the specific failure conditions and edge cases the team documented rather than what the developer summarized in the task

• Pull a card's checklist before generating implementation code: retrieve the checklist items the team agreed represent the implementation steps, so the generated code follows the order and scope the team defined rather than Codex's own interpretation of the task

• Read a feature card before scaffolding a module: fetch the card description that defines what the feature should do and what it excludes, so the generated scaffolding reflects the intended scope rather than a reconstruction from a brief summary

• Pull acceptance criteria comments before generating tests: retrieve the comment thread where the team refined what "done" looks like, so generated test cases cover the exact behaviors the team signed off on rather than generic happy-path cases

• Read a card's implementation notes before generating a migration script: fetch the comments where the team documented constraints, ordering requirements, and rollback considerations, so the generated script reflects the operational requirements the team identified

• Retrieve a blocked card's context before generating a workaround: pull the card and its activity to understand what was tried, what failed, and why it's blocked, so the generated alternative approach addresses the actual blocker rather than repeating a failed attempt

Why use Merge Agent Handler vs. a self-hosted Trello MCP server?

You can build a self-hosted Trello MCP server on top of Atlassian's REST API. The documentation is thorough, OAuth is well-supported, and for a single workspace the initial setup is manageable.

The problem is user and workspace scale. Trello OAuth tokens are scoped to individual user accounts.

If multiple developers run Codex tasks that need to read Trello cards, each requires a separate authorization flow, a separate token pair, and its own refresh cycle. A self-hosted server handles none of that, so credential storage and token refresh logic per user are still yours to manage.

Atlassian doesn't offer an official Trello MCP server, so there's no supported path beyond Merge or a community build. Merge Agent Handler handles multi-user OAuth centrally, scopes which Trello operations each agent can call, and logs every tool call with the agent identity and inputs.

For teams where Codex tasks read cards that contain product decisions or customer-specific context, that access control layer is worth having.

Why connect Trello to Codex?

Trello cards hold task context that doesn't travel cleanly into a Codex task description. Acceptance criteria defined in the card description, implementation steps documented in a checklist, edge cases flagged in a comment thread: those details exist in the card, and developers rarely copy all of them when creating a task.

When Codex only has the task description, it fills in the gaps. It generates code that addresses what the developer described and infers the rest.

When those inferences are wrong, the output looks correct until it hits the cases that were documented in the card but not in the task.

Connecting Trello gives Codex the ability to pull the card directly when a task references one. The generated fix covers the edge cases from the comment thread.

The generated scaffolding respects the constraints in the card description. The generated tests align with the acceptance criteria the team defined.

Can I use Merge Agent Handler's Trello MCP with my employees?

Yes, Agent Handler for Employees is built to help engineering organizations provision, secure, and govern how employees connect AI tools like Codex to project management tools like Trello.

Common patterns include:

• Provisioning and access control via SCIM with identity providers like Okta and Microsoft Entra ID, so IT can manage which employees can access which boards and card operations by role or team

• DLP and policy enforcement on tool calls, so admins can restrict which boards an employee's Codex agent can read and block retrieval of card content outside the employee's designated project scope before it reaches the agent's context

• User-level audit logging so security and IT teams can review which cards were accessed, which boards were queried, and which Trello data was retrieved, by which employee identity, and when

In practice, employees can use the Trello MCP to generate fixes grounded in real bug cards, scaffold features aligned with documented acceptance criteria, and produce scripts that reflect the implementation steps the team defined, while IT keeps centralized control over which boards and card data each developer's agent can reach.