Chain of Draft

Chain of Draft (CoD) MCP Server: An MCP server implementation of the Chain of Draft reasoning approach for more efficient LLM reasoning.

python

13

GitHub Stars

python

Language

6 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "stat-guy-chain-of-draft": {
      "command": "python3",
      "args": [
        "/absolute/path/to/cod/server.py"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

The Chain of Draft (CoD) MCP Server lets you run a fast, token-efficient reasoning workflow by generating concise intermediate steps and a final answer. It includes adaptive word limits, performance analytics, and open integration with the OpenAI API family, enabling you to solve diverse domain tasks with reduced cost and comparable accuracy.

How to use

You run the Chain of Draft MCP Server behind an MCP client or via the provided stdio configurations. You can solve math, code, logic, and domain-specific problems by sending a problem and optional domain to the server and receiving a final answer, short reasoning steps, and token usage.

How to install

Prerequisites you need to install and run the server are Python 3.10+ and Node.js 18+. You also need an Anthropic API key to access the underlying models.

# Python implementation prerequisites
python3 --version
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Python server run
python server.py

# JavaScript implementation prerequisites
node -v
npm install

# JavaScript server run
node index.js

Configuration and running

The server supports multiple ways to run. If you prefer a local, self-contained instance, you can start the Python or JavaScript server directly. If you integrate with your tools, you can point your MCP client to the appropriate stdio command as shown in the examples.

Security and environment

Keep your API keys secure. Use environment variables to pass credentials to the server, and avoid hard-coding secrets in scripts.

Troubleshooting

If the server fails to start, verify your Python or Node environments, confirm dependencies are installed, and ensure the ANTHROPIC_API_KEY (or equivalent) is set correctly in your environment.

Available tools

chain_of_draft_solve

Solve a problem using Chain of Draft reasoning and return a concise final answer with short reasoning steps.

math_solve

Solve mathematical problems with CoD, leveraging domain-specific adapters.

code_solve

Solve coding problems using CoD, returning concise reasoning and final answer.

logic_solve

Resolve logic problems with CoD, providing compact intermediate steps and result.

get_performance_stats

Retrieve performance statistics comparing CoD and CoT across domains.

get_token_reduction

Provide token reduction metrics achieved by CoD versus standard CoT.

analyze_problem_complexity

Analyze problem complexity to adjust word limits and adaptive behavior.