MCP Chain of Thought

MCP Chain of Thought provides an intelligent task management framework that helps you plan, decompose, and execute complex tasks with transparent reasoning and memory. It enables structured thought processes, task dependencies, and a configurable web interface to monitor progress across a project.

How to use

Set up your MCP client to talk to the chain-of-thought server as a local or remote service. You can initialize project rules to establish standards, plan tasks to create a development roadmap, review and adjust plans, then execute tasks and iterate in continuous mode if you want the agent to process all tasks sequentially. You’ll benefit from automatic task memory, a structured thought chain for step-by-step reasoning, and project-wide rules to maintain consistency across your work.

Typical workflow you can follow:

• Initialize project rules to establish guidelines for your project.
• Plan a task description to generate a development plan.
• Review the plan and provide feedback to refine the approach.
• Execute a specific task by name or ID.
• Switch to continuous mode to have tasks processed one after another until completion.

How to install

Prerequisites: you need Node.js and npm installed on your machine.

Via Smithery (recommended for quick setup):

npx -y @smithery/cli install @liorfranko/mcp-chain-of-thought --client claude

Manual installation (local development):

npm install
npm run build

Using with MCP-Compatible Clients

Configure your MCP client to connect to the chain-of-thought server. You will typically run the server as a local process or reference a remote endpoint if provided.

Below is an example configuration for a Cursor-like IDE client that runs the server locally via npx. This preserves the exact command pattern and environment variables you would use in practice.

{
  "mcpServers": {
    "chain_of_thought": {
      "command": "npx",
      "args": ["-y", "mcp-chain-of-thought"],
      "env": {
        "DATA_DIR": "/path/to/project/data",
        "ENABLE_THOUGHT_CHAIN": "true",
        "TEMPLATES_USE": "en",
        "ENABLE_GUI": "true",
        "ENABLE_DETAILED_MODE": "true"
      }
    }
  }
}

Configuration and environment

The server is configured via an MCP client configuration. The following environment variables are shown as part of the setup to control behavior and storage.

Key environment variables and their purposes:

Security and notes

Coordinate with your team to manage access to the chain-of-thought server. Use strong environment isolation and restrict access to the configured endpoint to prevent unauthorized task manipulation.

If you enable the GUI or detailed mode, ensure you have appropriate authentication and audit logging configured for sensitive project work.

Available tools

plan_task

Start planning tasks with structured descriptions and objectives.

analyze_task

Analyze requirements and constraints of a task.

process_thought

Perform step-by-step reasoning to approach a problem.

reflect_task

Improve solution concepts through reflection and iteration.

init_project_rules

Set project-wide standards and guidelines.

split_tasks

Decompose large tasks into subtasks.

list_tasks

Show all current tasks and statuses.

query_task

Search for specific tasks or details.

get_task_detail

Show detailed information about a task.

delete_task

Remove a task from the workflow.

execute_task

Run a specific task to produce results.

verify_task

Verify that a task meets requirements.

complete_task

Mark a task as completed.