Centralizing Requirements to Implementation: Using Rizm with Cursor via MCP

Keeping requirements, implementation plans, and results aligned is harder than it sounds.
Context gets lost between task trackers, wikis, and editors — especially once AI coding tools enter the workflow.

In this post, I’ll introduce a practical setup that connects Rizm (a self-hosted task + wiki tool) with the Cursor editor using MCP (Model Context Protocol), so that everything from planning to implementation stays task-centric and traceable.

---

What This Setup Solves

• No more copying requirements into chat prompts
• Plans and implementation notes are stored back with the task
• Tasks become the single source of truth — not your editor history

Cursor can read task data from Rizm and write back results automatically.

---

Prerequisites

• Rizm (self-hosted or demo environment)
• Cursor editor
• MCP support enabled in Cursor

---

Step 1: Generate an API Token in Rizm

From the Me page in Rizm, generate an API token for MCP access and copy it.

This token will be used by Cursor to securely access task data.

Step 2: Register Rizm as an MCP Server in Cursor

Open:
Cursor → Settings → Tools & MCP

Add your MCP server configuration:

"rizm": {
  "url": "https://demo.test-perspective.com/api/mcp",
  "headers": {
    "Authorization": "Bearer <YOUR_API_TOKEN>"
  }
}

Once registered, confirm that the server status shows as connected.

Step 3(optional): Define Rules in Cursor

In Cursor’s chat, use the /rules command to define how tasks should be handled.

Example:

For a given task key, fetch task details from Rizm and generate an implementation plan.

These rules allow Cursor to consistently interpret tasks without repeating instructions every time.

---

Workflow

1. Planning (Plan Mode)

After creating a working branch, switch Cursor to Plan Mode and run:

Create a plan for

Cursor retrieves the task description from Rizm via MCP and generates a plan based on both the requirements and the current codebase.

2. Implementation (Agent Mode)

Review the plan, then switch to Agent Mode to implement and test.

Because requirements are fetched directly from Rizm, there’s no need to paste task descriptions or keep switching tools.

3. Write Back to the Task

Once implementation is done, simply say:

Add the plan document as a comment

Cursor posts the plan back to the corresponding Rizm task as a comment — without changing task status or performing unrelated updates.

---

What Ends Up Stored in One Task

By the end of this flow, a single Rizm task contains:

• Original requirements
• Implementation plans
• References to branches and pull requests

This keeps historical context intact and makes later reviews much easier.

---

Why This Matters

AI-assisted development works best when context is stable and explicit.
By letting tasks — not chats — be the center of your workflow, you reduce ambiguity and improve long-term maintainability.

---

About Rizm

Rizm is a lightweight task management and wiki tool designed for self-hosted environments.
It combines the project-centric structure of tools like Jira or Backlog with a simpler, local-first approach.

Rizm is intended to run entirely within your own environment, giving you full control over your data and workflow.

• Demo site: https://demo.test-perspective.com/
• Distribution(GitHub): https://github.com/test-perspective/rizm-beta/blob/main/README.md

Comments

[ken_test-perspective]

beta7 released!

Improvements

• Enhanced user management and post-creation behavior.
• Improved branch and pull request workflow feedback.
• UI improvements across task, board, and issue detail views.
• Improved field editing experience, including date handling and layout flexibility.
• Minor usability improvements around SCM operations.

MCP / AIT

• Improved MCP API key setup experience.
• Enhanced AIT instruction capabilities.
• Known limitation: Manifest history cannot revert to the initial version.