Checking task status, reading requirements, and recording updates are routine parts of a project team’s day. After moving to a self-hosted project management platform, we looked at how AI integration could reduce that routine effort. We built a Plane MCP server that allows Claude to read and update work items in Plane for our project managers, so routine checks and status updates take place in a single conversation.
What Is MCP and How Does It Connect AI to Your Tools?
The Model Context Protocol (MCP) is an open standard introduced by Anthropic in November 2024. It gives AI assistants a common way to connect to the tools where a team’s work is recorded, so each tool needs one connector instead of a custom integration for every assistant.
Three parts make it work:
- Client: The AI application, such as Claude Desktop or Claude Code
- Server: A small program that exposes a tool’s features to the AI
- Tools: The specific actions the server offers, such as listing or updating tasks
In our setup, the flow is as follows:
Claude Desktop or Claude Code -> Plane MCP server (your computer) -> Plane (self-hosted)
The server receives a request from Claude, calls Plane’s API using the user’s own access key, and returns the result to the conversation.
What Does a Plane MCP Server Let AI Do?
Plane is an open-source project management platform built around work items, cycles, and modules. A Plane MCP server translates requests from Claude, the AI assistant developed by Anthropic, into Plane actions. Our version provides 77 core capabilities by default and 108 in extended mode (project documentation, 2026).
The AI can work across these areas:
- Projects: List, create, update, and delete projects, and view members
- Work items: List, search, create, update and delete, or open one by its identifier, such as DEV-42
- States and labels: Manage workflow states and labels
- Cycles and modules: Manage planning cycles and modules, including the work items inside them
- Epics, milestones and initiatives: Create and track larger bodies of work
- Comments, relations and links: Add comments, connect related work items and attach external links
- Activity and pages: Review a work item’s activity history, and create or read pages
- Bulk creation: Create many work items in one step, with automatic handling of request limits
Extended mode adds project and workspace administration, archives, custom work item types and properties, intake and time tracking.
Why Did We Move from ClickUp to Self-Hosted Plane?
Our project management team began in ClickUp and wanted more options than its free plan included. We looked for a platform that we could host ourselves, so that project data would remain on our own servers. We selected the open-source edition of Plane and now operate it in-house, consistent with our own mock API server, which we published as open source.
The move delivered three business benefits:
- Cost: Plane’s open-source edition runs on infrastructure we manage
- Data ownership: Project information stays on servers we control
- Independence: Delivery remains steady when a vendor changes its plans or pricing
Knowing exactly where project data is stored also supports conversations with clients in regulated sectors, such as banking.
Why Did We Build Our Own Plane MCP Server?
Hosting Plane internally provided the right foundation. The next step was to reduce how often people needed to open Plane only to check tasks and update statuses. Plane offers an official MCP server, and we built a version designed around our self-hosted setup, published as open source under the MIT License.
| Factor | Plane’s official server | Our server |
|---|---|---|
| Language | Python 3.10 or newer | Node.js 20 or newer |
| Start command | uvx | npx |
| Connection options | Local, or hosted over the web | Local |
| Sign-in | Browser sign-in (OAuth) or access key | Personal access key |
| Tool design | 30 tools covering 204 operations | 77 core tools, 108 in extended mode |
| Self-hosted Plane | Supported through a custom Plane address | Supported through a custom Plane address |
The two tool counts reflect different designs. The official server groups many operations under fewer tools, so the counts are best compared by design rather than by total. We keep our core set at 77 capabilities so that the assistant has fewer choices to weigh at each step.
Each option suits a different situation:
- Choose the official server for browser sign-in or hosted connections
- Choose our server for a lightweight setup that runs on each person’s own computer
- Choose either one for a self-hosted Plane instance, since both accept a custom Plane address
How Do Project Managers and Developers Use It Day to Day?
Project managers are the main users today, and developers can follow the same workflow. Each person runs the server on their own computer with their own Plane access key, and Plane remains the single record of every task.
A typical project manager workflow includes:
- Asking Claude, “Which work items in this project are still open?”
- Requesting the status and recent activity on a work item, such as DEV-42.
- Providing a client brief and asking Claude to create work items for each step.
- Adding a comment to DEV-42 with the day’s update.
For an individual task, the developer workflow is as follows:
- The developer asks Claude, “What tasks are assigned to me today?”
- Claude lists the tasks along with their deadlines and requirements.
- The developer selects the task Claude should work on.
- Claude reads the task requirements and assists with the work.
- The developer reviews the result, and once it is approved, Claude updates the task status in Plane.
The review in step 5 is intentional, because every status change reflects a person’s judgement.
How Do You Set Up the Plane MCP Server?
Setup is quick and requires Node.js 20 or newer. Each team member follows the same steps on their own computer.
- In Plane, open your account settings, go to API tokens and add a personal access token.
- Note your workspace name, which is the workspace part of your Plane address.
- Add the server to the Claude Desktop configuration file, using the settings below.
- Restart Claude Desktop and ask, “Show me all my Plane projects.”
{
“mcpServers”: {
“plane”: {
“command”: “npx”,
“args”: [“-y”, “@zethictech/plane-mcp-server”],
“env”: {
“PLANE_BASE_URL”: “https://your-plane-instance.com”,
“PLANE_API_KEY”: “YOUR_PLANE_ACCESS_KEY”,
“DEFAULT_WORKSPACE_SLUG”: “your-workspace-name”,
“PLANE_MCP_EXTENDED”: “false”
}
}
}
}
The settings work as follows:
- PLANE_BASE_URL: The address of your Plane instance (required)
- PLANE_API_KEY: Your personal access token from Plane (required)
- DEFAULT_WORKSPACE_SLUG: Your workspace name as it appears in the Plane address (required)
- PLANE_MCP_EXTENDED: Set to “true” to enable all 108 capabilities (optional)
Claude Code users can register the same server with the claude mcp add command and the same settings. If the server needs attention, these checks resolve most cases:
- Confirm that all three required settings are present
- Remove any trailing slash from the Plane address
- Run npx @zethictech/plane-mcp-server –version to confirm the package loads
- Refresh the cached copy by running npx –yes @zethictech/plane-mcp-server
- Confirm that the access key can reach the workspace
What Has Changed for Us, and How Does It Improve Productivity?
The benefits of this AI integration appear in routine activity, particularly for project managers. The table compares the earlier process with the current one.
| Activity | Earlier process | With the Plane MCP server |
|---|---|---|
| Checking tasks, deadlines and status | Opening Plane and filtering manually | Asking Claude in plain language |
| Reading requirements | Moving between several tools | Claude reads the task and works from it |
| Updating status | A manual update after each task | An update after the task is reviewed |
| Routine project follow-up | Regular checks on progress | Attention on decisions and reviews |
| Task tracking | Maintained manually in Plane | Maintained through the conversation, still in Plane |
Different roles benefit in different ways:
- Project managers: Current task status, deadlines and updates in one conversation, with more time for decisions and reviews
- Developers: A clear view of the day’s work and its requirements in one conversation
- Engineering leads: Task records stay in Plane, where the whole team can find them
How Can AI Update Project Tasks Responsibly?
Reliability comes from how the workflow is designed. We keep a person involved at the point where a task status changes. This is a core consideration in agentic AI development, where assistants act inside business systems.
- The server runs on each user’s own computer, with no third-party connector between Claude and Plane
- Task details are shared with Claude during a conversation, so teams in regulated sectors should review this against their data policy
- Each person connects with their own Plane access key, so activity is tied to their account
- The access key is stored in the user’s private settings
- The extended set of 108 capabilities is enabled by choice, with the 77 core capabilities as the default
Teams with additional security requirements can add regular key rotation and access reviews.
What Helps Teams Get Better Results?
Claude works from the content of each task, so the following practices improve results:
- Write task requirements clearly, with acceptance criteria where possible
- Use work item identifiers such as DEV-42 in prompts for precise answers
- Begin with read-only questions while the team builds confidence
- Pilot the server with one project before extending it across the workspace
How Can Your Team Apply the Same Approach?
We built this server to address a requirement within our own team. The same approach guides our client work: identify the routine step that takes up a team’s time, then connect AI to the system where the work is already recorded. Because the server is open source, other teams using Plane can install it and adapt it to their needs.
This approach is part of our AI and engineering services:
- AI workflow automation for routine, repeatable tasks
- Agentic AI development for assistants that act inside business systems
- Custom application development for integrations with self-hosted platforms
Plan a Similar Integration for Your Team
At Zethic, we use a Plane MCP server to bring routine task tracking into a conversation, using the tools our team already has. Teams that track work in a self-hosted tool can follow the same approach. Our engineering team can review an existing workflow and suggest a practical integration. We recommend starting with read-only access to one project, so the team can see how it works before any updates are enabled. Further access can then be added in stages as the team gains confidence.