dev-mcp
A local MCP server that gives AI agents access to developer tooling — GitHub (read-only), documentation search, and web research — via stdio transport.
README
dev-mcp
A local MCP (Model Context Protocol) server that gives AI agents access to developer tooling — GitHub (read-only), documentation search, and web research — via stdio transport.
Built with Bun, TypeScript, the MCP SDK, and Zod.
Features
GitHub (read-only)
| Tool | Description |
|---|---|
github.repo_search |
Search repositories by query |
github.repo_read |
Get repository details |
github.file_read |
Read a file from a repository |
github.issues_list |
List issues with state/label filters |
github.issue_get |
Get a single issue |
github.prs_list |
List pull requests |
github.pr_diff |
Get the diff of a pull request |
github.commit_search |
Search commits by message or author |
Context7 Documentation Search
| Tool | Description |
|---|---|
context7.docs_search |
Search library documentation |
context7.docs_get |
Get specific documentation page |
context7.examples_get |
Get code examples |
context7.api_lookup |
Look up API symbol documentation |
Exa Web Research
| Tool | Description |
|---|---|
exa.web_search |
Search the web |
exa.web_fetch |
Fetch and convert a URL to markdown/text |
exa.code_search |
Search code across GitHub |
exa.research_brief |
Generate a research brief by synthesizing top results |
Total: 16 tools
Prerequisites
- Bun 1.3+ (runtime)
- A GitHub Personal Access Token with
repo(read-only) andpublic_reposcopes
Quick Start
# Install dependencies
bun install
# Create your environment file
cp .env.example .env
# Edit .env and set GITHUB_TOKEN
# Run typecheck
bun run typecheck
# Start the server
bun start
The server listens on stdio, ready to receive MCP JSON-RPC messages.
Configuration
All configuration is via environment variables:
| Variable | Required | Default | Description |
|---|---|---|---|
GITHUB_TOKEN |
Yes | — | GitHub PAT (read-only access) |
CONTEXT7_API_KEY |
No | — | Context7 API key |
EXA_API_KEY |
No | — | Exa API key |
SENTRY_DSN |
No | — | Sentry DSN for error reporting |
SENTRY_ENVIRONMENT |
No | development |
Sentry environment tag |
LOG_LEVEL |
No | info |
Log level (debug, info, warn, error) |
NODE_ENV |
No | development |
Runtime environment |
Architecture
src/
├── config/ # Environment validation (Zod schema)
├── core/ # MCP server, tool registry, transport
├── observability/ # Logger, Sentry integration
├── schemas/ # Zod schemas for all tool inputs
├── security/ # Input validation wrapper
├── tools/
│ ├── context7/ # Documentation search tools
│ ├── exa/ # Web search & research tools
│ └── github/ # GitHub API tools
└── utils/ # Error hierarchy (McpError)
Development
bun run typecheck # Type-check all files
bun start # Start the server (stdio)
The project uses Husky for pre-commit hooks and Turbo for task orchestration. A typecheck hook runs automatically before each commit.
License
MIT — see LICENSE.
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。
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。