claude-101

27 AI tools as MCP server + CLI + Skill. Just talk to Claude — it handles the rest.

English | 繁體中文 | 日本語

What is this?

Install once, get 24 superpowers. Claude 101 is an MCP server that gives Claude real computation abilities — statistics, code analysis, SQL parsing, financial math, and more — things LLMs cannot do reliably on their own.

How it works:

You: "Compare React, Vue, and Svelte for our project"
  ↓
Skill tells Claude to call build_comparison_matrix
  ↓
MCP tool computes: Vue 8.1 > React 7.9 > Svelte 7.5 (weighted scoring)
  ↓
Claude writes: "Vue leads by 0.2 points. The result is sensitive to
               the DX weight — if you value Ecosystem more, React wins."

Without claude-101: Claude guesses at numbers and rankings. With claude-101: Claude uses precise computation, then reasons about the results.

Setup (2 minutes)

Step 1: Add MCP Server

Add to your .mcp.json (project root or ~/.claude/.mcp.json):

{
  "mcpServers": {
    "claude-101": {
      "command": "uvx",
      "args": ["--from", "claude-101[mcp]", "claude-101-server"]
    }
  }
}

Step 2: Install Skill

The Skill teaches Claude when to call each tool and how to use every field in the result:

mkdir -p ~/.claude/skills/claude-101-mastery/references
cd ~/.claude/skills/claude-101-mastery
BASE=https://raw.githubusercontent.com/claude-world/claude-101/main/skills/claude-101-mastery
curl -sLO $BASE/SKILL.md
curl -sLO $BASE/references/writing-workflows.md   --output-dir references
curl -sLO $BASE/references/analysis-workflows.md   --output-dir references
curl -sLO $BASE/references/coding-workflows.md     --output-dir references
curl -sLO $BASE/references/business-workflows.md   --output-dir references

Done. Start a new Claude Code session and just talk naturally.

24 Use Cases

After setup, you can simply ask Claude to do any of these — the Skill handles the rest:

Writing & Communication

Analysis & Research

Coding & Technical

Business & Productivity

CLI Usage

Also works as a standalone command-line tool:

# Install
pip install "claude-101[mcp]"

# List all tools
claude-101 list
claude-101 list --category analysis

# Run any tool directly
claude-101 draft-email "meeting follow-up" --tone assertive
claude-101 --pretty analyze-data "name,score\nAlice,95\nBob,87"
claude-101 scaffold-proposal business "Cloud Migration" --investment 100000 --annual-return 50000

# Pipe from stdin
echo "SELECT * FROM users" | claude-101 process-sql -
cat mycode.py | claude-101 analyze-code -

# Tool help
claude-101 draft-email --help

Python Library

from claude_101.analysis.data import analyze_data
from claude_101.business.decision import evaluate_decision

result = analyze_data("name,score\nAlice,95\nBob,87", output_format="csv", operations="all")
result["correlations"]  # [{"column_a": "score", "column_b": "hours", "pearson_r": 0.94}]

result = evaluate_decision("A,B", "Speed,Cost", "0.6,0.4", "A:Speed=9,Cost=5;B:Speed=6,Cost=9")
result["winner"]  # {"option": "A", "score": 7.4, "margin": 0.2}

Architecture

claude-101/
  src/claude_101/
    server.py           # MCP server (27 tools via FastMCP)
    cli.py              # CLI (auto-generated from function signatures)
    _utils.py           # 14 shared computation functions
    _guides.py          # 24 embedded use-case guides
    writing/            # 6 tools: email, blog, meeting, social, techdoc, story
    analysis/           # 6 tools: data, summary, comparison, survey, financial, legal
    coding/             # 6 tools: codegen, review, sql, apidoc, testgen, adr
    business/           # 6 tools: planning, interview, proposal, support, prd, decision
  skills/
    claude-101-mastery.md  # Skill file (teaches Claude how to use all 24 tools)
  tests/                   # 157 tests across 6 files

Dependencies: Only sqlparse (everything else is stdlib). MCP is optional.

Contributing

See CONTRIBUTING.md for guidelines. See CHANGELOG.md for release history.

License

MIT — see LICENSE.