mcp-to-skill

将任何 MCP 服务器封装为 Claude Skill，支持 stdio/SSE/HTTP 传输协议，提供 CLI 和 Python SDK 双模式。

Features

• ✅ 双模式支持: CLI 命令行工具 + Python SDK
• ✅ 多传输协议: stdio/SSE/HTTP
• ✅ uv 依赖管理: 比 pip 快 10-100 倍
• ✅ 自动 introspect: 自动获取工具列表
• ✅ 一键转换: 从配置到可用技能
• ✅ 统计追踪: 工具调用统计和日志
• ✅ 上下文节省: 96%+ 上下文节省

Installation

Python SDK (pip)

pip install mcp-to-skill

Bun CLI (本地)

# 确保已安装 uv
curl -LsSf https://astral.sh/uv/install.sh | sh

# 使用本地 lib.ts
bun lib.ts convert my-mcp.json

Usage

Python SDK

from mcp_to_skill import (
    MCPConfig,
    SkillConfig,
    Transport,
    convert_to_skill,
    validate_skill,
    test_skill,
    get_skill_status
)

# Convert MCP server to skill
config = MCPConfig(
    name="github",
    transport=Transport.STDIO,
    command="npx",
    args=["@modelcontextprotocol/server-github"],
    env={"GITHUB_TOKEN": "your-token"}
)

skill_info = convert_to_skill(config)
print(f"Skill created at: {skill_info.path}")

# Validate skill
validation = validate_skill(skill_info.path)
print(f"Valid: {validation['valid']}")
print(f"Tools: {len(validation['tools'])}")

# Get status
status = get_skill_status(skill_info.path)
print(f"Total calls: {status['stats']['total_calls']}")

CLI

# Convert
mcp-to-skill convert my-mcp.json

# Validate
mcp-to-skill validate ~/.claude/skills/my-mcp

# Test
mcp-to-skill test ~/.claude/skills/my-mcp --mode list

# Status
mcp-to-skill status ~/.claude/skills/my-mcp

# Reset stats
mcp-to-skill reset-stats ~/.claude/skills/my-mcp

Bun CLI (本地)

# Convert
bun lib.ts convert my-mcp.json

# Validate
bun lib.ts validate ~/.claude/skills/my-mcp

# Test
bun lib.ts test ~/.claude/skills/my-mcp --list

MCP Config Format

{
  "name": "my-mcp",
  "transport": "stdio|sse|http",
  "command": "npx",  // stdio only
  "args": ["@example/mcp-server"],  // stdio only
  "endpoint": "https://...",  // sse/http only
  "env": {"API_KEY": "your-key"},
  "keep_alive": {
    "enabled": true,
    "timeout": 3600
  }
}

Examples

Example 1: GitHub MCP (stdio)

from mcp_to_skill import MCPConfig, Transport, convert_to_skill

config = MCPConfig(
    name="github",
    transport=Transport.STDIO,
    command="npx",
    args=["-y", "@modelcontextprotocol/server-github"],
    env={"GITHUB_TOKEN": "ghp_your_token"}
)

skill = convert_to_skill(config)

Example 2: DeepWiki (SSE)

from mcp_to_skill import MCPConfig, Transport, convert_to_skill

config = MCPConfig(
    name="deepwiki",
    transport=Transport.SSE,
    endpoint="https://mcp.deepwiki.com/sse"
)

skill = convert_to_skill(config)

Example 3: Enable Process Reuse

from mcp_to_skill import MCPConfig, Transport, convert_to_skill

config = MCPConfig(
    name="my-mcp",
    transport=Transport.STDIO,
    command="npx",
    args=["@example/mcp-server"],
    keep_alive={
        "enabled": True,
        "timeout": 3600,
        "check_interval": 60
    }
)

skill = convert_to_skill(config)

API Reference

MCPConfig

@dataclass
class MCPConfig:
    name: str
    transport: Transport = Transport.STDIO
    command: Optional[str] = None
    args: Optional[List[str]] = None
    endpoint: Optional[str] = None
    env: Optional[Dict[str, str]] = None
    keep_alive: Optional[Dict[str, Any]] = None

convert_to_skill()

def convert_to_skill(
    mcp_config: MCPConfig,
    skill_config: Optional[SkillConfig] = None
) -> SkillInfo

validate_skill()

def validate_skill(skill_path: str) -> Dict[str, Any]

test_skill()

def test_skill(
    skill_path: str,
    mode: Literal["list", "describe", "call"] = "list",
    tool_name: Optional[str] = None,
    arguments: Optional[Dict[str, Any]] = None
) -> Dict[str, Any]

get_skill_status()

def get_skill_status(skill_path: str) -> Dict[str, Any]

reset_skill_stats()

def reset_skill_stats(skill_path: str) -> Dict[str, Any]

Performance

Context Savings

Tools	MCP	Skill	Saved
8	4000 tokens	150 tokens	96%
20	10000 tokens	150 tokens	98.5%

Dependency Management

Metric	pip	uv	Improvement
Install time	10s+	<1s	10x+
Virtual env	Manual	Auto	✅
Resolution	Slow	Fast	5x+

Process Reuse

Scenario	Without Reuse	With Reuse	Improvement
First call	5s	5s	-
Subsequent calls	5s	<0.5s	10x
10 calls	50s	5s	10x

Requirements

Python SDK

• Python 3.10+
• httpx>=0.25.0

Bun CLI

• Bun runtime
• uv (https://astral.sh/uv)

Documentation

• Python SDK Guide
• MCP 规范
• uv 文档

License

MIT

Credits

Based on mcp-to-skill-converter by GBSOSS