GitHub Integration Hub
Enables AI agents to interact with GitHub through OAuth-authenticated operations including starting authorization flows, listing repositories, and creating issues using stored access tokens.
README
GitHub Integration Hub
Built by Cristian to demonstrate OAuth orchestration, secure token storage, SDK/CLI tooling, and MCP servers in TypeScript.
Why this project exists
I wanted a hands-on showcase proving I can:
- Design an OAuth flow (GitHub) with state handling and token persistence.
- Build a lightweight data layer on SQLite using TypeScript.
- Expose the same business logic via HTTP routes, an SDK, a CLI, and an MCP server for AI agents.
Everything runs on Node.js LTS with Express, better-sqlite3, and TypeScript.
Feature highlights
- (OAuth)
/auth/github/startand/auth/github/callbackwrap the GitHub dance. - (Storage) Automatic SQLite table creation plus refresh scaffolding and repository helpers.
- (Services) Shared orchestration reused by HTTP routes, the CLI, the SDK, and MCP tools.
- (Tooling) Type-safe
IntegrationHubClientSDK together with thegithub-integration-hubCLI. - (Agents) MCP tools
github_start_auth,github_list_repos, andgithub_create_issue.
Getting started
1. Install dependencies
npm install
2. Configure environment
Copy .env.example to .env and set your GitHub OAuth credentials:
PORT=3000
DATABASE_FILE=./github-integration-hub.db
GITHUB_CLIENT_ID=your-client-id
GITHUB_CLIENT_SECRET=your-client-secret
GITHUB_OAUTH_AUTHORIZE_URL=https://github.com/login/oauth/authorize
GITHUB_OAUTH_TOKEN_URL=https://github.com/login/oauth/access_token
GITHUB_API_BASE_URL=https://api.github.com
GITHUB_REDIRECT_URI=http://localhost:3000/auth/github/callback
GITHUB_INTEGRATION_HUB_BASE_URL=http://localhost:3000
Ensure the GitHub OAuth app redirect matches
GITHUB_REDIRECT_URI.
3. Run the backend
Development mode with hot reload:
npm run dev
Production-style build + run:
npm run build
npm start
The server logs: GitHub Integration Hub listening on port <port>.
MCP server (for AI agents)
Compile once, then launch the stdio server:
npm run build
npm run start:mcp
Tools available:
github_start_auth– returns an authorization URL for auserId.github_list_repos– lists repos for the stored connection.github_create_issue– posts an issue using the saved token.
Point your MCP-aware IDE/CLI (for example Codex) to this stdio process and the tools become callable immediately.
HTTP API reference
| Method | Path | Description |
|---|---|---|
| POST | /auth/github/start |
Returns { authUrl } for userId |
| GET | /auth/github/callback |
Exchanges code + state, saves tokens |
| GET | /connections |
Lists connections (?userId= optional) |
| POST | /actions/github/list-repos |
Lists repos via stored token |
| POST | /actions/github/create-issue |
Creates an issue |
| GET | /health |
Service heartbeat |
CLI (github-integration-hub)
npm run build
npx github-integration-hub connections:list
npx github-integration-hub actions:list-repos --user alice
npx github-integration-hub actions:create-issue --user alice --owner my-org --repo api --title "Hello" --body "Proof from Cristian"
Flags:
--base-urloverridesGITHUB_INTEGRATION_HUB_BASE_URL.connections:list --user <userId>filters entries.
SDK example
import { IntegrationHubClient } from 'github-integration-hub/sdk';
const client = new IntegrationHubClient({ baseUrl: process.env.GITHUB_INTEGRATION_HUB_BASE_URL! });
async function showcase() {
const { authUrl } = await client.startGithubAuth('user-123');
console.log('Send someone to authorize:', authUrl);
const repos = await client.listGithubRepos('user-123');
console.log(repos.map((repo) => repo.full_name));
}
Manual OAuth test plan
POST /auth/github/startwith{ "userId": "demo-user" }.- Visit the returned
authUrl, approve the GitHub app. - Callback stores the connection automatically.
- Trigger actions using HTTP, the CLI, SDK, or MCP tools.
Project structure
.
|-- cli/ # CLI entry point
|-- sdk/ # TypeScript SDK
|-- src/
| |-- config/ # Environment helpers
| |-- db/ # SQLite init + repository
| |-- mcp/ # MCP server
| |-- providers/ # GitHub OAuth + API clients
| |-- routes/ # Express routers
| `-- services/ # Token manager + actions
|-- .env.example
|-- package.json
|-- tsconfig.json
`-- README.md
Made by Cristian. Ping me if you want to talk about backend integrations, OAuth design, or MCP tooling.
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。