skill-seekers-reference/docs/MCP_SETUP.md

Complete MCP Setup Guide - MCP 2025 (v2.4.0)

Step-by-step guide to set up the Skill Seeker MCP server with 5 supported AI coding agents.

Version 2.4.0 Highlights:

• ✅ MCP SDK v1.25.0 - Latest protocol support (upgraded from v1.18.0)
• ✅ FastMCP Framework - Modern, decorator-based server implementation
• ✅ Dual Transport - HTTP + stdio support (choose based on agent)
• ✅ 17 MCP Tools - Expanded from 9 tools (8 new source management tools)
• ✅ Multi-Agent Support - Claude Code, Cursor, Windsurf, VS Code + Cline, IntelliJ IDEA
• ✅ Auto-Configuration - One-line setup with ./setup_mcp.sh
• ✅ Production Ready - 34 comprehensive tests, 100% pass rate

---

Table of Contents

• What's New in v2.4.0
• Migration from v2.3.0
• Prerequisites
• Quick Start (Recommended)
• Manual Installation
• Agent-Specific Configuration
• Transport Modes
• Verification
• Usage Examples
• Troubleshooting
• Advanced Configuration

---

What's New in v2.4.0

MCP 2025 Upgrade

MCP SDK v1.25.0 (upgraded from v1.18.0):

• Latest MCP protocol specification
• Enhanced reliability and performance
• Better error handling and diagnostics

FastMCP Framework:

• Decorator-based tool registration (modern Python pattern)
• Simplified server implementation (2200 lines → 708 lines, 68% reduction)
• Modular tool architecture in tools/ directory
• Easier to maintain and extend

Dual Transport Support:

• stdio transport: Default, backward compatible with Claude Code and VS Code + Cline
• HTTP transport: New, required for Cursor, Windsurf, and IntelliJ IDEA
• Automatic transport detection via agent_detector.py

New Features

17 MCP Tools (expanded from 9):

Config Tools (3):

• generate_config - Generate config for any documentation site
• list_configs - List all available preset configurations
• validate_config - Validate config file structure

Scraping Tools (4):

• estimate_pages - Estimate page count before scraping
• scrape_docs - Scrape documentation and build skill
• scrape_github - Scrape GitHub repositories
• scrape_pdf - Extract content from PDF files

Packaging Tools (3):

• package_skill - Package skill into .zip file
• upload_skill - Upload .zip to Claude AI (NEW)
• install_skill - Install skill to AI coding agents (NEW)

Splitting Tools (2):

• split_config - Split large documentation configs
• generate_router - Generate router/hub skills

Source Tools (5 - NEW):

• fetch_config - Fetch configs from API or git sources
• submit_config - Submit new configs to community
• add_config_source - Register private git repositories as config sources
• list_config_sources - List all registered config sources
• remove_config_source - Remove registered config sources

Multi-Agent Support:

• 5 supported agents with automatic detection
• Auto-configuration script (./setup_mcp.sh) detects and configures all agents
• Transport auto-selection based on agent requirements

Infrastructure

HTTP Server Features:

• Health check endpoint: http://localhost:8000/health
• SSE endpoint: http://localhost:8000/sse
• Configurable host and port
• Production-ready with uvicorn

New Server Implementation:

• server_fastmcp.py - New FastMCP-based server (recommended)
• server.py - Legacy server (deprecated, maintained for compatibility)

---

Migration from v2.3.0

If you're upgrading from v2.3.0, follow these steps:

1. Update Dependencies

# Navigate to repository
cd /path/to/Skill_Seekers

# Update package
pip install -e . --upgrade

# Verify MCP SDK version
python3 -c "import mcp; print(mcp.__version__)"
# Should show: 1.25.0 or higher

2. Update Configuration

For Claude Code (no changes required):

{
  "mcpServers": {
    "skill-seeker": {
      "command": "python",
      "args": ["-m", "skill_seekers.mcp.server_fastmcp"]
    }
  }
}

For HTTP-based agents (Cursor, Windsurf, IntelliJ):

Old config (v2.3.0):

{
  "command": "python",
  "args": ["-m", "skill_seekers.mcp.server", "--http", "--port", "3000"]
}

New config (v2.4.0):

{
  "url": "http://localhost:3000/sse"
}

The HTTP server now runs separately and agents connect via URL instead of spawning the server.

3. Start HTTP Server (if using HTTP agents)

# Start HTTP server on port 3000
python -m skill_seekers.mcp.server_fastmcp --http --port 3000

# Or use custom host/port
python -m skill_seekers.mcp.server_fastmcp --http --host 0.0.0.0 --port 8080

4. Test Configuration

In any connected agent:

List all available MCP tools

You should see 17 tools (up from 9 in v2.3.0).

5. Optional: Run Auto-Configuration

The easiest way to update all agents:

./setup_mcp.sh

This will:

• Detect all installed agents
• Configure stdio agents (Claude Code, VS Code + Cline)
• Show HTTP server setup instructions for HTTP agents (Cursor, Windsurf, IntelliJ)

---

Prerequisites

Required Software

1. Python 3.10 or higher

   python3 --version
   # Should show: Python 3.10.x or higher
   

2. AI Coding Agent (at least one):

   • Claude Code - Download from claude.ai/code
   • Cursor - Download from cursor.sh
   • Windsurf - Download from codeium.com/windsurf
   • VS Code + Cline - Install Cline extension
   • IntelliJ IDEA - Download from jetbrains.com

3. Skill Seeker repository (for source installation):

   git clone https://github.com/yusufkaraaslan/Skill_Seekers.git
   cd Skill_Seekers
   

   Or install from PyPI:

   pip install skill-seekers
   

System Requirements

• Operating System: macOS, Linux, or Windows (WSL)
• Disk Space: 100 MB for dependencies + space for generated skills
• Network: Internet connection for documentation scraping

---

Quick Start (Recommended)

The fastest way to set up MCP for all detected agents:

1. Run Auto-Configuration Script

# Navigate to repository
cd /path/to/Skill_Seekers

# Run setup script
./setup_mcp.sh

2. What the Script Does

1. Detects Python version - Ensures Python 3.10+
2. Installs dependencies - Installs MCP SDK v1.25.0, FastMCP, uvicorn
3. Detects agents - Automatically finds installed AI coding agents
4. Configures stdio agents - Auto-configures Claude Code and VS Code + Cline
5. Shows HTTP setup - Provides commands for Cursor, Windsurf, IntelliJ IDEA

3. Follow On-Screen Instructions

For stdio agents (Claude Code, VS Code + Cline):

• Restart the agent
• Configuration is automatic

For HTTP agents (Cursor, Windsurf, IntelliJ):

• Start HTTP server: python -m skill_seekers.mcp.server_fastmcp --http --port 3000
• Add server URL to agent settings (instructions provided by script)
• Restart the agent

4. Verify Setup

In your agent:

List all available MCP tools

You should see 17 Skill Seeker tools.

---

Manual Installation

If you prefer manual setup or the auto-configuration script doesn't work:

Step 1: Install Python Dependencies

# Navigate to repository root
cd /path/to/Skill_Seekers

# Install package in editable mode (includes all dependencies)
pip install -e .

# Or install specific dependencies manually
pip install "mcp>=1.25,<2" requests beautifulsoup4 uvicorn

Expected output:

Successfully installed mcp-1.25.0 fastmcp-... uvicorn-... requests-2.31.0 beautifulsoup4-4.12.3

Step 2: Verify Installation

# Test stdio mode
timeout 3 python3 -m skill_seekers.mcp.server_fastmcp || echo "Server OK (timeout expected)"

# Test HTTP mode
python3 -c "import uvicorn; print('HTTP support available')"

Step 3: Note Your Repository Path

# Get absolute path
pwd

# Example output: /Users/username/Projects/Skill_Seekers
# or: /home/username/Skill_Seekers

Save this path - you'll need it for configuration!

---

Agent-Specific Configuration

Claude Code (stdio transport)

Config Location:

• macOS: ~/Library/Application Support/Claude/mcp.json
• Linux: ~/.config/claude-code/mcp.json
• Windows: %APPDATA%\Claude\mcp.json

Configuration:

{
  "mcpServers": {
    "skill-seeker": {
      "command": "python",
      "args": ["-m", "skill_seekers.mcp.server_fastmcp"]
    }
  }
}

With custom Python path:

{
  "mcpServers": {
    "skill-seeker": {
      "command": "/usr/local/bin/python3.11",
      "args": ["-m", "skill_seekers.mcp.server_fastmcp"]
    }
  }
}

Setup Steps:

1. Create config directory: mkdir -p ~/Library/Application\ Support/Claude
2. Edit config: nano ~/Library/Application\ Support/Claude/mcp.json
3. Paste configuration above
4. Save and exit
5. Restart Claude Code

---

Cursor (HTTP transport)

Config Location:

• macOS: ~/Library/Application Support/Cursor/mcp_settings.json
• Linux: ~/.cursor/mcp_settings.json
• Windows: %APPDATA%\Cursor\mcp_settings.json

Step 1: Start HTTP Server

# Terminal 1 - Run HTTP server
python -m skill_seekers.mcp.server_fastmcp --http --port 3000

# Should show:
# INFO: Started server process
# INFO: Uvicorn running on http://127.0.0.1:3000

Step 2: Configure Cursor

{
  "mcpServers": {
    "skill-seeker": {
      "url": "http://localhost:3000/sse"
    }
  }
}

Step 3: Verify Connection

# Check health endpoint
curl http://localhost:3000/health

# Should return: {"status": "ok"}

Step 4: Restart Cursor

---

Windsurf (HTTP transport)

Config Location:

• macOS: ~/Library/Application Support/Windsurf/mcp_config.json
• Linux: ~/.windsurf/mcp_config.json
• Windows: %APPDATA%\Windsurf\mcp_config.json

Step 1: Start HTTP Server

# Terminal 1 - Run HTTP server
python -m skill_seekers.mcp.server_fastmcp --http --port 3001

# Use different port if Cursor is using 3000

Step 2: Configure Windsurf

{
  "mcpServers": {
    "skill-seeker": {
      "url": "http://localhost:3001/sse"
    }
  }
}

Step 3: Restart Windsurf

---

VS Code + Cline Extension (stdio transport)

Config Location:

• macOS: ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json
• Linux: ~/.config/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json
• Windows: %APPDATA%\Code\User\globalStorage\saoudrizwan.claude-dev\settings\cline_mcp_settings.json

Configuration:

{
  "mcpServers": {
    "skill-seeker": {
      "command": "python",
      "args": ["-m", "skill_seekers.mcp.server_fastmcp"]
    }
  }
}

Setup Steps:

1. Install Cline extension in VS Code
2. Open Cline settings (Cmd/Ctrl + Shift + P → "Cline: Settings")
3. Navigate to MCP settings
4. Add configuration above
5. Reload VS Code window

---

IntelliJ IDEA (HTTP transport)

Config Location:

• macOS: ~/Library/Application Support/JetBrains/IntelliJIdea2024.3/mcp.xml
• Linux: ~/.config/JetBrains/IntelliJIdea2024.3/mcp.xml
• Windows: %APPDATA%\JetBrains\IntelliJIdea2024.3\mcp.xml

Step 1: Start HTTP Server

# Terminal 1 - Run HTTP server
python -m skill_seekers.mcp.server_fastmcp --http --port 3002

Step 2: Configure IntelliJ

Edit mcp.xml:

<?xml version="1.0" encoding="UTF-8"?>
<application>
  <component name="MCPSettings">
    <servers>
      <server>
        <name>skill-seeker</name>
        <url>http://localhost:3002/sse</url>
      </server>
    </servers>
  </component>
</application>

Step 3: Restart IntelliJ IDEA

---

Transport Modes

stdio Transport (Default)

How it works:

• Agent spawns MCP server as subprocess
• Communication via stdin/stdout
• Server lifecycle managed by agent

Advantages:

• Automatic process management
• No port conflicts
• Zero configuration after setup

Supported Agents:

• Claude Code
• VS Code + Cline

Usage:

{
  "command": "python",
  "args": ["-m", "skill_seekers.mcp.server_fastmcp"]
}

No additional steps needed - agent handles everything.

---

HTTP Transport (New)

How it works:

• MCP server runs as HTTP server
• Agents connect via SSE (Server-Sent Events)
• Single server can support multiple agents

Advantages:

• Multiple agents can share one server
• Easier debugging (can test with curl)
• Production-ready with uvicorn

Supported Agents:

• Cursor
• Windsurf
• IntelliJ IDEA

Usage:

Step 1: Start HTTP Server

# Default (port 8000)
python -m skill_seekers.mcp.server_fastmcp --http

# Custom port
python -m skill_seekers.mcp.server_fastmcp --http --port 3000

# Custom host and port
python -m skill_seekers.mcp.server_fastmcp --http --host 0.0.0.0 --port 8080

# Debug mode
python -m skill_seekers.mcp.server_fastmcp --http --log-level DEBUG

Step 2: Configure Agent

{
  "url": "http://localhost:8000/sse"
}

Step 3: Test Endpoints

# Health check
curl http://localhost:8000/health
# Returns: {"status": "ok"}

# SSE endpoint (agent connects here)
curl http://localhost:8000/sse
# Returns SSE stream

---

Verification

Step 1: Check MCP Server Loaded

In your AI coding agent, type:

List all available MCP tools

You should see 17 Skill Seeker tools:

Config Tools:

• generate_config - Generate config for documentation site
• list_configs - List available preset configs
• validate_config - Validate config structure

Scraping Tools:

• estimate_pages - Estimate page count
• scrape_docs - Scrape documentation
• scrape_github - Scrape GitHub repositories
• scrape_pdf - Extract PDF content

Packaging Tools:

• package_skill - Package skill into .zip
• upload_skill - Upload to Claude AI
• install_skill - Install to AI agents

Splitting Tools:

• split_config - Split large configs
• generate_router - Generate router skills

Source Tools:

• fetch_config - Fetch configs from sources
• submit_config - Submit new configs
• add_config_source - Register git sources
• list_config_sources - List config sources
• remove_config_source - Remove sources

Step 2: Test a Simple Command

List all available configs

Expected response:

Available configurations:
1. godot - Godot Engine documentation
2. react - React framework
3. vue - Vue.js framework
4. django - Django web framework
5. fastapi - FastAPI Python framework
6. kubernetes - Kubernetes documentation
7. steam-economy-complete - Steam Economy API
... (24 total configs)

Step 3: Test Config Generation

Generate a config for Tailwind CSS at https://tailwindcss.com/docs

Expected response:

✅ Config created: configs/tailwind.json

Step 4: Test HTTP Server (if using)

# Health check
curl http://localhost:8000/health

# Should return:
{"status": "ok"}

# Check SSE endpoint
curl -N http://localhost:8000/sse

# Should stream SSE events

---

Usage Examples

Example 1: Generate Skill from Scratch

User: Generate config for Svelte docs at https://svelte.dev/docs

Agent: ✅ Config created: configs/svelte.json

User: Estimate pages for configs/svelte.json

Agent: 📊 Estimated pages: 150
       Recommended max_pages: 180

User: Scrape docs using configs/svelte.json

Agent: ✅ Skill created at output/svelte/

User: Package skill at output/svelte/

Agent: ✅ Created: output/svelte.zip
       Ready to upload to Claude!

Example 2: Use Private Config Repository

User: Add config source named "team" with git URL https://github.com/myorg/skill-configs.git

Agent: ✅ Source registered: team
       Type: github
       Priority: 100

User: List config sources

Agent: Registered sources:
       1. team (github, enabled, priority: 100)
       2. official (api, enabled, priority: 200)

User: Fetch config "nextjs" from source "team"

Agent: ✅ Config downloaded: configs/nextjs.json
       Source: team (https://github.com/myorg/skill-configs.git)

User: Scrape docs using configs/nextjs.json

Agent: [Scraping starts...]

Example 3: Multi-Source Workflow

User: List config sources

Agent: No sources registered. Use add_config_source to add sources.

User: Add config source "company-internal" with git URL git@gitlab.company.com:configs/ai-skills.git

Agent: ✅ Source registered: company-internal
       Type: gitlab
       Token: GITLAB_TOKEN (environment variable)

User: Fetch config "internal-api" from "company-internal"

Agent: ✅ Config downloaded: configs/internal-api.json

User: Validate configs/internal-api.json

Agent: ✅ Config is valid
       - Base URL: https://docs.company.com/api/
       - Max pages: 1000
       - Categories: 5

User: Scrape docs using configs/internal-api.json

Agent: [Scraping internal documentation...]

---

Troubleshooting

Issue: MCP Server Not Loading

Symptoms:

• Skill Seeker tools don't appear in agent
• No response when asking about configs

Solutions:

1. Check configuration file exists:

   # Claude Code
   cat ~/Library/Application\ Support/Claude/mcp.json
   
   # Cursor
   cat ~/Library/Application\ Support/Cursor/mcp_settings.json
   

2. Verify Python path:

   which python3
   # Should show: /usr/bin/python3 or similar
   

3. Test server manually:

   For stdio:

   timeout 3 python3 -m skill_seekers.mcp.server_fastmcp
   # Should exit cleanly or timeout (both OK)
   

   For HTTP:

   python3 -m skill_seekers.mcp.server_fastmcp --http --port 8000
   # Should show: Uvicorn running on http://127.0.0.1:8000
   

4. Check agent logs:

   Claude Code:

   • macOS: ~/Library/Logs/Claude/
   • Linux: ~/.config/claude-code/logs/

   Cursor:

   • macOS: ~/Library/Logs/Cursor/
   • Linux: ~/.cursor/logs/

5. Completely restart agent:

   • Quit agent (don't just close window)
   • Kill any background processes: pkill -f skill_seekers
   • Reopen agent

---

Issue: "ModuleNotFoundError: No module named 'mcp'"

Solution:

# Install package
pip install -e .

# Or install dependencies manually
pip install "mcp>=1.25,<2" requests beautifulsoup4 uvicorn

Verify installation:

python3 -c "import mcp; print(mcp.__version__)"
# Should show: 1.25.0 or higher

---

Issue: HTTP Server Not Starting

Symptoms:

• python -m skill_seekers.mcp.server_fastmcp --http fails
• "ModuleNotFoundError: No module named 'uvicorn'"

Solution:

# Install uvicorn
pip install uvicorn

# Or install with extras
pip install -e ".[mcp]"

Verify uvicorn:

python3 -c "import uvicorn; print('OK')"

---

Issue: Port Already in Use

Symptoms:

• "Address already in use" when starting HTTP server

Solution:

# Find process using port
lsof -i :8000

# Kill process
kill -9 <PID>

# Or use different port
python -m skill_seekers.mcp.server_fastmcp --http --port 8001

---

Issue: Tools Appear But Don't Work

Symptoms:

• Tools listed but commands fail
• "Error executing tool" messages

Solutions:

1. Check working directory:

   For stdio agents, ensure package is installed:

   pip install -e .
   

2. Verify CLI tools exist:

   python3 -m skill_seekers.cli.doc_scraper --help
   python3 -m skill_seekers.cli.package_skill --help
   

3. Test tool directly:

   # Test in Python
   python3 -c "from skill_seekers.mcp.tools import list_configs_impl; print('OK')"
   

4. Check HTTP server logs (if using HTTP transport):

   python -m skill_seekers.mcp.server_fastmcp --http --log-level DEBUG
   

---

Issue: Agent Can't Connect to HTTP Server

Symptoms:

• Agent shows connection error
• curl to /health fails

Solutions:

1. Verify server is running:

   curl http://localhost:8000/health
   # Should return: {"status": "ok"}
   

2. Check firewall:

   # macOS
   sudo /usr/libexec/ApplicationFirewall/socketfilterfw --getglobalstate
   
   # Linux
   sudo ufw status
   

3. Test with different host:

   # Try 0.0.0.0 instead of 127.0.0.1
   python -m skill_seekers.mcp.server_fastmcp --http --host 0.0.0.0
   

4. Check agent config URL:

   {
     "url": "http://localhost:8000/sse"  // Not /health!
   }
   

---

Issue: Slow or Hanging Operations

Solutions:

1. Check rate limit in config:

   • Default: 0.5 seconds
   • Increase if needed: 1.0 or 2.0 seconds

2. Use smaller max_pages for testing:

   Generate config with max_pages=20 for testing
   

3. Check network connection:

   curl -I https://docs.example.com
   

4. Enable debug logging:

   python -m skill_seekers.mcp.server_fastmcp --http --log-level DEBUG
   

---

Advanced Configuration

Custom Environment Variables

For stdio agents:

{
  "mcpServers": {
    "skill-seeker": {
      "command": "python",
      "args": ["-m", "skill_seekers.mcp.server_fastmcp"],
      "env": {
        "ANTHROPIC_API_KEY": "sk-ant-...",
        "GITHUB_TOKEN": "ghp_...",
        "GITLAB_TOKEN": "glpat-...",
        "PYTHONPATH": "/custom/path"
      }
    }
  }
}

For HTTP server:

# Set environment variables before starting
export ANTHROPIC_API_KEY=sk-ant-...
export GITHUB_TOKEN=ghp_...
python -m skill_seekers.mcp.server_fastmcp --http

---

Multiple Python Versions

If you have multiple Python versions:

Find Python path:

which python3.11
# /usr/local/bin/python3.11

Use in config:

{
  "mcpServers": {
    "skill-seeker": {
      "command": "/usr/local/bin/python3.11",
      "args": ["-m", "skill_seekers.mcp.server_fastmcp"]
    }
  }
}

---

Virtual Environment

To use a Python virtual environment:

# Create venv
cd /path/to/Skill_Seekers
python3 -m venv venv
source venv/bin/activate

# Install package
pip install -e .

# Get Python path
which python3
# Copy this path

Use in config:

{
  "mcpServers": {
    "skill-seeker": {
      "command": "/path/to/Skill_Seekers/venv/bin/python3",
      "args": ["-m", "skill_seekers.mcp.server_fastmcp"]
    }
  }
}

---

Running HTTP Server as Service

systemd (Linux):

Create /etc/systemd/system/skill-seeker-mcp.service:

[Unit]
Description=Skill Seeker MCP HTTP Server
After=network.target

[Service]
Type=simple
User=yourusername
WorkingDirectory=/path/to/Skill_Seekers
ExecStart=/usr/bin/python3 -m skill_seekers.mcp.server_fastmcp --http --port 8000
Restart=on-failure
Environment="ANTHROPIC_API_KEY=sk-ant-..."

[Install]
WantedBy=multi-user.target

Enable and start:

sudo systemctl enable skill-seeker-mcp
sudo systemctl start skill-seeker-mcp
sudo systemctl status skill-seeker-mcp

macOS (launchd):

Create ~/Library/LaunchAgents/com.skillseeker.mcp.plist:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
    <key>Label</key>
    <string>com.skillseeker.mcp</string>
    <key>ProgramArguments</key>
    <array>
        <string>/usr/local/bin/python3</string>
        <string>-m</string>
        <string>skill_seekers.mcp.server_fastmcp</string>
        <string>--http</string>
        <string>--port</string>
        <string>8000</string>
    </array>
    <key>WorkingDirectory</key>
    <string>/path/to/Skill_Seekers</string>
    <key>RunAtLoad</key>
    <true/>
    <key>KeepAlive</key>
    <true/>
    <key>StandardOutPath</key>
    <string>/tmp/skill-seeker-mcp.log</string>
    <key>StandardErrorPath</key>
    <string>/tmp/skill-seeker-mcp.error.log</string>
</dict>
</plist>

Load:

launchctl load ~/Library/LaunchAgents/com.skillseeker.mcp.plist
launchctl start com.skillseeker.mcp

---

Debug Mode

Enable verbose logging for troubleshooting:

stdio transport:

{
  "mcpServers": {
    "skill-seeker": {
      "command": "python",
      "args": [
        "-u",
        "-m",
        "skill_seekers.mcp.server_fastmcp"
      ],
      "env": {
        "DEBUG": "1"
      }
    }
  }
}

HTTP transport:

python -m skill_seekers.mcp.server_fastmcp --http --log-level DEBUG

---

Complete Example Configurations

Minimal (Recommended for Most Users)

Claude Code (stdio):

{
  "mcpServers": {
    "skill-seeker": {
      "command": "python",
      "args": ["-m", "skill_seekers.mcp.server_fastmcp"]
    }
  }
}

Cursor (HTTP):

Start server:

python -m skill_seekers.mcp.server_fastmcp --http --port 3000

Config:

{
  "mcpServers": {
    "skill-seeker": {
      "url": "http://localhost:3000/sse"
    }
  }
}

---

With API Keys and Custom Tokens

Claude Code:

{
  "mcpServers": {
    "skill-seeker": {
      "command": "python",
      "args": ["-m", "skill_seekers.mcp.server_fastmcp"],
      "env": {
        "ANTHROPIC_API_KEY": "sk-ant-your-key-here",
        "GITHUB_TOKEN": "ghp_your-token-here"
      }
    }
  }
}

HTTP Server:

export ANTHROPIC_API_KEY=sk-ant-your-key-here
export GITHUB_TOKEN=ghp_your-token-here
python -m skill_seekers.mcp.server_fastmcp --http --port 3000

---

Multiple Agents Sharing HTTP Server

Start one HTTP server:

python -m skill_seekers.mcp.server_fastmcp --http --port 8000

Configure all HTTP agents to use it:

Cursor (~/Library/Application Support/Cursor/mcp_settings.json):

{
  "mcpServers": {
    "skill-seeker": {
      "url": "http://localhost:8000/sse"
    }
  }
}

Windsurf (~/Library/Application Support/Windsurf/mcp_config.json):

{
  "mcpServers": {
    "skill-seeker": {
      "url": "http://localhost:8000/sse"
    }
  }
}

IntelliJ (~/Library/Application Support/JetBrains/IntelliJIdea2024.3/mcp.xml):

<component name="MCPSettings">
  <servers>
    <server>
      <name>skill-seeker</name>
      <url>http://localhost:8000/sse</url>
    </server>
  </servers>
</component>

All three agents now share the same MCP server instance!

---

End-to-End Workflow

Complete Setup and First Skill

# 1. Install from source
cd ~/Projects
git clone https://github.com/yusufkaraaslan/Skill_Seekers.git
cd Skill_Seekers

# 2. Run auto-configuration
./setup_mcp.sh

# 3. Follow prompts
# - Installs dependencies
# - Detects agents
# - Configures automatically

# 4. For HTTP agents, start server
python -m skill_seekers.mcp.server_fastmcp --http --port 3000

# 5. Restart your AI coding agent

# 6. Test in agent:

In your agent:

User: List all available configs
User: Scrape docs using configs/react.json with max 50 pages
User: Package skill at output/react/

Result: output/react.zip ready to upload!

---

Next Steps

After successful setup:

1. Try preset configs:

   • React: scrape docs using configs/react.json
   • Vue: scrape docs using configs/vue.json
   • Django: scrape docs using configs/django.json

2. Create custom configs:

   • generate config for [framework] at [url]

3. Set up private config sources:

   • add config source "team" with git URL https://github.com/myorg/configs.git

4. Test with small limits first:

   • Use max_pages parameter: scrape docs using configs/test.json with max 20 pages

5. Explore enhancement:

   • Use --enhance-local flag for AI-powered SKILL.md improvement

---

Getting Help

• Documentation:

  • README.md - User guide
  • CLAUDE.md - Technical architecture
  • ENHANCEMENT.md - Enhancement guide
  • UPLOAD_GUIDE.md - Upload instructions

• Issues: GitHub Issues

• Agent Detection: See agent_detector.py

• Auto-Configuration: See setup_mcp.sh

---

Quick Reference Card

SETUP:
1. Install: pip install -e .
2. Configure: ./setup_mcp.sh
3. Restart agent

VERIFY:
- "List all available MCP tools" (should show 17 tools)
- "List all available configs" (should show 24 configs)

GENERATE SKILL:
1. "Generate config for [name] at [url]"
2. "Estimate pages for configs/[name].json"
3. "Scrape docs using configs/[name].json"
4. "Package skill at output/[name]/"

PRIVATE CONFIGS:
1. "Add config source [name] with git URL [url]"
2. "List config sources"
3. "Fetch config [name] from [source]"

TRANSPORT MODES:
- stdio: Claude Code, VS Code + Cline (automatic)
- HTTP: Cursor, Windsurf, IntelliJ (requires server)

START HTTP SERVER:
python -m skill_seekers.mcp.server_fastmcp --http --port 3000

TROUBLESHOOTING:
- Check: cat ~/.config/claude-code/mcp.json
- Test stdio: timeout 3 python -m skill_seekers.mcp.server_fastmcp
- Test HTTP: curl http://localhost:8000/health
- Logs (Claude Code): ~/Library/Logs/Claude/
- Kill servers: pkill -f skill_seekers

---

Happy skill creating! 🚀