toolcall
Enables creating MCP (Model Context Protocol) servers with zero boilerplate, full TypeScript support, and multiple transports (stdio and HTTP).
README
toolcall
Create MCP (Model Context Protocol) servers with zero boilerplate.
Features
- Minimal API - Just two functions:
serve()andtool() - Type-safe - Full TypeScript support with Zod schema validation
- Multiple transports - Supports both stdio and HTTP
- MCP compliant - Implements MCP protocol version
2024-11-05 - Client included - Connect to any MCP server programmatically
Installation
npm install toolcall zod
Quick Start
Create an MCP server in just a few lines:
import { serve, tool } from 'toolcall'
import { z } from 'zod'
serve({
name: 'my-server',
version: '1.0.0',
tools: {
greet: tool({
description: 'Greet someone by name',
parameters: z.object({
name: z.string().describe('The name of the person to greet')
}),
execute: ({ name }) => `Hello, ${name}!`
}),
add: tool({
description: 'Add two numbers',
parameters: z.object({
a: z.number().describe('First number'),
b: z.number().describe('Second number')
}),
execute: ({ a, b }) => ({ result: a + b })
})
}
})
Run it:
npx tsx server.ts
Claude Code Integration
toolcall servers integrate seamlessly with Claude Code. Add your server to Claude Code's MCP configuration:
1. Create your server file
// my-tools.ts
import { serve, tool } from 'toolcall'
import { z } from 'zod'
serve({
name: 'my-tools',
tools: {
get_weather: tool({
description: 'Get current weather for a city',
parameters: z.object({
city: z.string().describe('City name'),
unit: z.enum(['celsius', 'fahrenheit']).default('celsius')
}),
execute: async ({ city, unit }) => {
// Your implementation here
return { city, temperature: 22, unit, condition: 'sunny' }
}
})
}
})
2. Configure Claude Code
Add to your Claude Code MCP settings (~/.claude/claude_desktop_config.json or via Claude Code settings):
{
"mcpServers": {
"my-tools": {
"command": "npx",
"args": ["tsx", "/path/to/my-tools.ts"]
}
}
}
Or if you've compiled your TypeScript:
{
"mcpServers": {
"my-tools": {
"command": "node",
"args": ["/path/to/my-tools.js"]
}
}
}
3. Use in Claude Code
Once configured, Claude Code will automatically discover your tools. You can ask Claude to use them:
"Use my get_weather tool to check the weather in Tokyo"
API Reference
serve(options)
Creates and starts an MCP server.
serve({
name: 'my-server', // Server name (default: 'toolcall-server')
version: '1.0.0', // Server version (default: '1.0.0')
transport: 'stdio', // Transport type: 'stdio' | 'http' (default: 'stdio')
port: 3000, // Port for HTTP transport (default: 3000)
tools: { // Tool definitions
// ... your tools
}
})
tool(definition)
Defines a type-safe tool with Zod schema validation.
tool({
description: 'Tool description shown to clients',
parameters: z.object({
// Zod schema for parameters
}),
execute: async (params) => {
// Tool implementation
// Can return string, object, or any JSON-serializable value
}
})
Parameter Types
toolcall supports all Zod types:
import { z } from 'zod'
// Strings
z.string()
z.string().min(1).max(100)
z.string().email()
z.string().url()
// Numbers
z.number()
z.number().min(0).max(100)
z.number().int()
// Booleans
z.boolean()
// Enums
z.enum(['option1', 'option2', 'option3'])
// Arrays
z.array(z.string())
// Optional with defaults
z.string().optional()
z.number().default(10)
// Descriptions (shown in tool schema)
z.string().describe('Parameter description')
Return Values
Tools can return any JSON-serializable value:
// String return
execute: ({ name }) => `Hello, ${name}!`
// Object return (automatically JSON-stringified)
execute: ({ a, b }) => ({ result: a + b, operation: 'addition' })
// Async operations
execute: async ({ url }) => {
const response = await fetch(url)
return await response.json()
}
Transports
Stdio (Default)
The stdio transport reads JSON-RPC messages from stdin and writes responses to stdout. This is the standard transport for MCP servers used by Claude Code and other MCP clients.
serve({
transport: 'stdio', // or omit - stdio is default
tools: { /* ... */ }
})
HTTP
The HTTP transport creates an HTTP server that accepts JSON-RPC POST requests.
serve({
transport: 'http',
port: 3000,
tools: { /* ... */ }
})
Test with curl:
# Initialize
curl -X POST http://localhost:3000 \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{}}}'
# List tools
curl -X POST http://localhost:3000 \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}'
# Call a tool
curl -X POST http://localhost:3000 \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"greet","arguments":{"name":"World"}}}'
Client Usage
toolcall includes a client for connecting to any MCP server:
import { connect } from 'toolcall'
// Connect to a stdio server
const client = await connect('npx tsx ./server.ts')
// Or connect to an HTTP server
const client = await connect('http://localhost:3000')
// List available tools
console.log(client.listTools())
// Call a tool
const result = await client.call('greet', { name: 'World' })
console.log(result) // "Hello, World!"
// Clean up
client.close()
Complete Example
import { serve, tool } from 'toolcall'
import { z } from 'zod'
serve({
name: 'example-server',
version: '1.0.0',
tools: {
// Simple string return
greet: tool({
description: 'Greet someone by name',
parameters: z.object({
name: z.string().describe('The name of the person to greet')
}),
execute: ({ name }) => `Hello, ${name}!`
}),
// Object return
add: tool({
description: 'Add two numbers together',
parameters: z.object({
a: z.number().describe('First number'),
b: z.number().describe('Second number')
}),
execute: ({ a, b }) => ({ result: a + b })
}),
// Async with enum and default
get_weather: tool({
description: 'Get the current weather for a city',
parameters: z.object({
city: z.string().describe('City name'),
unit: z.enum(['celsius', 'fahrenheit']).default('celsius').describe('Temperature unit')
}),
execute: async ({ city, unit }) => {
// Simulate API call
const temp = Math.round(Math.random() * 30 + 10)
const tempInUnit = unit === 'fahrenheit' ? Math.round(temp * 9 / 5 + 32) : temp
return {
city,
temperature: tempInUnit,
unit,
condition: ['sunny', 'cloudy', 'rainy'][Math.floor(Math.random() * 3)]
}
}
}),
// Constrained parameters
search: tool({
description: 'Search for information',
parameters: z.object({
query: z.string().describe('Search query'),
limit: z.number().min(1).max(100).default(10).describe('Maximum results')
}),
execute: async ({ query, limit }) => {
return {
query,
results: Array.from({ length: Math.min(limit, 3) }, (_, i) => ({
title: `Result ${i + 1} for "${query}"`,
url: `https://example.com/result/${i + 1}`
}))
}
}
})
}
})
Error Handling
toolcall automatically validates parameters against your Zod schemas. Invalid parameters return a JSON-RPC error:
{
"jsonrpc": "2.0",
"id": 1,
"error": {
"code": -32602,
"message": "Invalid parameters",
"data": {
"name": { "_errors": ["Required"] }
}
}
}
Errors thrown in tool execution are caught and returned as internal errors:
{
"jsonrpc": "2.0",
"id": 1,
"error": {
"code": -32603,
"message": "Internal error",
"data": "Error message here"
}
}
Protocol Details
toolcall implements the Model Context Protocol specification:
- Protocol Version:
2024-11-05 - Transport: JSON-RPC 2.0 over stdio or HTTP
- Methods:
initialize- Server initialization handshakenotifications/initialized- Client initialization acknowledgmenttools/list- List available toolstools/call- Execute a toolping- Health check
Development
# Install dependencies
npm install
# Build
npm run build
# Watch mode
npm run dev
# Run example server
npx tsx examples/server.ts
# Run tests
npm test
License
MIT
Author
Yi Min Yang (https://www.yiminyang.dev)
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。