@spec2tools/stdio-mcp
Exposes any OpenAPI spec endpoints as AI agent tools via stdio, requiring no code generation or maintenance.
README
Spec2Tools
Dynamically convert OpenAPI specs into AI agent tools at runtime—no code generation required.
Packages
This monorepo contains four packages:
| Package | Description | Install |
|---|---|---|
| @spec2tools/core | Core utilities for OpenAPI parsing and authentication | npm install @spec2tools/core |
| @spec2tools/sdk | Create AI SDK tools from OpenAPI specifications | npm install @spec2tools/sdk |
| @spec2tools/cli | CLI for interacting with OpenAPI-based AI tools | npx @spec2tools/cli |
| @spec2tools/stdio-mcp | MCP server exposing OpenAPI endpoints as tools via stdio | npx @spec2tools/stdio-mcp |
Quick Start
Using the SDK
import { createTools } from '@spec2tools/sdk';
import { generateText, stepCountIs } from 'ai';
import { openai } from '@ai-sdk/openai';
const tools = await createTools({ spec: './openapi.yaml' });
const result = await generateText({
model: openai('gpt-4o'),
tools,
prompt: 'List all users',
stopWhen: stepCountIs(3)
});
console.log(result.text);
Using the CLI
# Set your OpenAI API key
export OPENAI_API_KEY=your-api-key
# Start the agent with an OpenAPI spec
npx @spec2tools/cli start --spec https://api.example.com/openapi.json
Using as MCP Server
# Add to Claude Code (or any MCP client)
claude mcp add --transport stdio my-api \
-- npx @spec2tools/stdio-mcp ./path/to/openapi.yaml
Code Mode
Collapse all API endpoints into just 2 tools (search + execute), cutting input token usage by ~99.9%:
# MCP server with code mode
claude mcp add --transport stdio my-api \
-- npx @spec2tools/stdio-mcp ./openapi.yaml --code-mode
// SDK with code mode
const tools = await createTools({
spec: './openapi.yaml',
codeMode: true,
});
You can also convert any existing AI SDK tools to code mode:
import { toCodeModeTools } from '@spec2tools/core';
const codeModeTools = toCodeModeTools(existingTools);
Thesis
You have a server. You have an OpenAPI spec. Now, with all the MCP hype, you're told you need to build another server—an MCP server—just so AI agents can call your API. You look it up: define tools, maintain handlers, keep it in sync with your actual API. But wait... you already described your API in OpenAPI. Why do we need to describe it to the agent twice?
Spec2Tools is the alternative. It reads your OpenAPI spec and exposes endpoints as tools—no code generation, no maintenance burden. Use it directly with the AI SDK, or spin up an MCP server instantly from your existing spec.
Development
# Clone the repo
git clone https://github.com/CahidArda/spec2tools.git
cd spec2tools
# Install dependencies
pnpm install
# Build all packages
pnpm build
License
MIT
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。