analyze_problem_complexity
Analyze the complexity of a problem
LABSCO SUMMARY
It restructures the prompt before your model sees it, so reasoning arrives as short drafted steps instead of an open-ended chain.
The trade is explicit: it calls out to a second model, which means another key and another round trip, and in return you cap how many words each reasoning step may use. The two stats tools report how Chain of Draft has actually compared with Chain of Thought on the problems this server has run — check them before deciding it earns a slot.
What it is
A reasoning wrapper: your prompt is transformed into a structured drafting format, passed to whichever LLM you have configured, and the response is transformed back into a concise answer.
What you get
- chain_of_draft_solve for any reasoning problem, with a domain hint, a choice between the Chain of Draft and Chain of Thought approach, a cap on words per reasoning step, and an adaptive word limit
- Three specialised entry points carrying the same controls: math_solve, code_solve and logic_solve
- analyze_problem_complexity, to size a problem up before deciding how much reasoning to spend on it
- get_performance_stats and get_token_reduction, which report how the two approaches have compared on the problems this server has run
- Bring your own model: Anthropic Claude, OpenAI, Mistral, any Ollama model, or a local chat-completion endpoint
Requirements
A model of your own and the credentials for it: ANTHROPIC_API_KEY, OPENAI_API_KEY or MISTRAL_API_KEY. For a local model, set MCP_LLM_PROVIDER=ollama with MCP_OLLAMA_MODEL, or MCP_LLM_PROVIDER=custom with MCP_CUSTOM_LLM_ENDPOINT pointing at your own chat-completion API. The repository ships two implementations — Python 3.10+ and Node.js 22+ — and the npm package is mcp-chain-of-draft-prompt-tool.
Setup effort
One command plus a key — npx -y mcp-chain-of-draft-prompt-tool, then supply credentials
USE CASES
01
Working a hard problem in short, bounded steps
Send the problem with a domain hint and a per-step word cap instead of an open-ended prompt.
02
Getting a maths answer with the working visible
Route the problem through the maths entry point so each step is drafted and checkable.
03
Reasoning through a coding problem before writing any code
Draft the approach step by step, then take the result into the editor.
04
Untangling a logic puzzle
Use the logic entry point, which applies the same drafting format to constraint-style problems.
05
Deciding how much reasoning a problem deserves
Score its complexity first, then pick the approach and the word limit to match.
06
Checking whether the shorter format is paying off
Read the recorded comparison between the two approaches and the token difference between them.
7 tools
Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.
chain_of_draft_solve
Solve a reasoning problem using Chain of Draft approach
code_solve
Solve a coding problem using Chain of Draft reasoning
get_performance_stats
Get performance statistics for CoD vs CoT approaches
get_token_reduction
Get token reduction statistics for CoD vs CoT
logic_solve
Solve a logic problem using Chain of Draft reasoning
math_solve
Solve a math problem using Chain of Draft reasoning
BEFORE YOU INSTALL
What to know first
Tools you get
7 tools when we connected. Four tools are the same solver with a domain hint (chain_of_draft_solve, math_solve, code_solve, logic_solve), and two report statistics.
Requires
- Node.js 22+ for the npx and JavaScript routes; Python 3.10+ only for the Python route
- One model provider key (ANTHROPIC_API_KEY, OPENAI_API_KEY or MISTRAL_API_KEY), or a local Ollama model
CONTEXT COST
Light Standing cost
Rated by the size of the 7 tool definitions that ride along on every turn, not by how many there are
INSTALLATION
Connect your client
Server setup
TERMINAL
npx -y mcp-chain-of-draft-prompt-toolMCP Clients
Claude Desktop — npx
CLAUDE DESKTOP — NPX · JSON
{
"mcpServers": {
"chain-of-draft-prompt-tool": {
"command": "npx",
"args": [
"-y",
"mcp-chain-of-draft-prompt-tool"
],
"env": {
"ANTHROPIC_API_KEY": "<YOUR_ANTHROPIC_API_KEY>"
}
}
}
}Runs the published npm package (mcp-chain-of-draft-prompt-tool), so there is nothing to clone or build.
The blocks set ANTHROPIC_API_KEY. To use OpenAI or Mistral instead, put OPENAI_API_KEY or MISTRAL_API_KEY in the block's env section in its place; for a local Ollama model, set MCP_LLM_PROVIDER to ollama and MCP_OLLAMA_MODEL to the model you pulled.
Claude Desktop — Python
CLAUDE DESKTOP — PYTHON · JSON
{
"mcpServers": {
"chain-of-draft-prompt-tool": {
"command": "python3",
"args": [
"<YOUR_PATH_TO_COD>/server.py"
],
"env": {
"ANTHROPIC_API_KEY": "<YOUR_ANTHROPIC_API_KEY>"
}
}
}
}This route runs the Python implementation from a clone: git clone https://github.com/brendancopley/mcp-chain-of-draft-prompt-tool, then pip install -r requirements.txt inside it (Python 3.10+). <YOUR_PATH_TO_COD>/server.py is the full path of server.py in that clone.
The blocks set ANTHROPIC_API_KEY. To use OpenAI or Mistral instead, put OPENAI_API_KEY or MISTRAL_API_KEY in the block's env section in its place; for a local Ollama model, set MCP_LLM_PROVIDER to ollama and MCP_OLLAMA_MODEL to the model you pulled.
Claude Desktop — JavaScript
CLAUDE DESKTOP — JAVASCRIPT · JSON
{
"mcpServers": {
"chain-of-draft-prompt-tool": {
"command": "node",
"args": [
"<YOUR_PATH_TO_COD>/index.js"
],
"env": {
"ANTHROPIC_API_KEY": "<YOUR_ANTHROPIC_API_KEY>"
}
}
}
}This route runs the JavaScript build from a clone: git clone https://github.com/brendancopley/mcp-chain-of-draft-prompt-tool, then npm install and npm run nx build inside it (Node.js 22+). <YOUR_PATH_TO_COD>/index.js is the full path of dist/index.js in that clone (npm start runs the compiled server from dist).
The blocks set ANTHROPIC_API_KEY. To use OpenAI or Mistral instead, put OPENAI_API_KEY or MISTRAL_API_KEY in the block's env section in its place; for a local Ollama model, set MCP_LLM_PROVIDER to ollama and MCP_OLLAMA_MODEL to the model you pulled.
Claude Desktop — single executable
CLAUDE DESKTOP — SINGLE EXECUTABLE · JSON
{
"mcpServers": {
"chain-of-draft-prompt-tool": {
"command": "<YOUR_PATH_TO_MCP_CHAIN_OF_DRAFT>",
"env": {
"ANTHROPIC_API_KEY": "<YOUR_ANTHROPIC_API_KEY>"
}
}
}
}This route runs a standalone executable you build yourself: in a clone of the repository, npm install, then npm run build:sea (or build:macos, build:linux, build:windows). <YOUR_PATH_TO_MCP_CHAIN_OF_DRAFT> is the full path of the executable it writes to dist.
The blocks set ANTHROPIC_API_KEY. To use OpenAI or Mistral instead, put OPENAI_API_KEY or MISTRAL_API_KEY in the block's env section in its place; for a local Ollama model, set MCP_LLM_PROVIDER to ollama and MCP_OLLAMA_MODEL to the model you pulled.
Claude Code — Python
CLAUDE CODE — PYTHON · TERMINAL
claude mcp add chain-of-draft-prompt-tool -e ANTHROPIC_API_KEY=<YOUR_ANTHROPIC_API_KEY> -- python3 <YOUR_PATH_TO_COD>/server.pyThis route runs the Python implementation from a clone: git clone https://github.com/brendancopley/mcp-chain-of-draft-prompt-tool, then pip install -r requirements.txt inside it (Python 3.10+). <YOUR_PATH_TO_COD>/server.py is the full path of server.py in that clone.
The blocks set ANTHROPIC_API_KEY. To use OpenAI or Mistral instead, put OPENAI_API_KEY or MISTRAL_API_KEY in the block's env section in its place; for a local Ollama model, set MCP_LLM_PROVIDER to ollama and MCP_OLLAMA_MODEL to the model you pulled.
Claude Code — JavaScript
CLAUDE CODE — JAVASCRIPT · TERMINAL
claude mcp add chain-of-draft-prompt-tool -e ANTHROPIC_API_KEY=<YOUR_ANTHROPIC_API_KEY> -- node <YOUR_PATH_TO_COD>/index.jsThis route runs the JavaScript build from a clone: git clone https://github.com/brendancopley/mcp-chain-of-draft-prompt-tool, then npm install and npm run nx build inside it (Node.js 22+). <YOUR_PATH_TO_COD>/index.js is the full path of dist/index.js in that clone (npm start runs the compiled server from dist).
The blocks set ANTHROPIC_API_KEY. To use OpenAI or Mistral instead, put OPENAI_API_KEY or MISTRAL_API_KEY in the block's env section in its place; for a local Ollama model, set MCP_LLM_PROVIDER to ollama and MCP_OLLAMA_MODEL to the model you pulled.
MCP Inspector — test
MCP INSPECTOR — TEST · TERMINAL
npx @modelcontextprotocol/inspector -e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY -- node dist/index.jsThe blocks set ANTHROPIC_API_KEY. To use OpenAI or Mistral instead, put OPENAI_API_KEY or MISTRAL_API_KEY in the block's env section in its place; for a local Ollama model, set MCP_LLM_PROVIDER to ollama and MCP_OLLAMA_MODEL to the model you pulled.
