Sourcebook
@maroondlabs
关于 Sourcebook
Live codebase intelligence for AI agents. Analyzes import graphs (PageRank), git history (reverted commits, co-change coupling), and conventions to generate context files (CLAUDE.md, .cursorrules, copilot-instructions.md). Only surfaces non-discoverable information — things agent
基本信息
配置
使用下面的配置,将此服务器添加到你的 MCP 客户端。
{
"mcpServers": {
"sourcebook": {
"command": "npx",
"args": [
"-y",
"sourcebook",
"serve"
]
}
}
}工具
未检测到工具
工具是从 README 中自动提取的。维护者可以在 ## Tools 标题下列出工具,即可填充这部分内容。
概览
What is Sourcebook?
Sourcebook is a safety layer for code changes that analyzes git diffs for completeness — it flags files that should have been modified but weren't. It uses rules-based structural detection plus optional AI-powered semantic analysis, with a claim of zero false positives on clean diffs. It is designed for developers and AI agents working in version-controlled repositories.
How to use Sourcebook?
Sourcebook can be used via CLI, as a Claude Code hook, or as an MCP server. For MCP usage, run npx sourcebook serve and add the server configuration to your MCP client. To analyze a diff, run npx sourcebook check with optional flags like --ai (requires ANTHROPIC_API_KEY), --quiet (exit code only), --json, or --branch <branch>. Run npx sourcebook init to set up Claude Code hooks and generate context files.
Key features of Sourcebook
- Detects missing test files when a source file changes
- Identifies sibling modules that import or are imported by changed files
- Flags co-change companions based on historical git commit patterns
- Warns about hub file blast radius (files with many dependents)
- AI analysis catches semantic relationships like field renames needing migrations
- Supports TypeScript/JavaScript, Python, Go, and Rust with full import graph analysis
Use cases of Sourcebook
- Catching missing test updates when a source file is modified
- Ensuring sibling modules and related configuration files are updated together
- Scanning recent commits for previously missed file changes (
scan-history) - Using as a pre-commit hook via Claude Code integration to prevent incomplete commits
FAQ from Sourcebook
What does Sourcebook catch?
Sourcebook detects missing test files, unchanged sibling modules, files that historically co-change, and hub files with wide blast radius. With the --ai flag, it also finds cross-module semantic relationships, field renames needing migrations, and stale validation logic.
Does Sourcebook require an AI model?
No. The rules-based analysis (co-change, test detection, import graph, hub detection) works without an LLM and runs in under one second. The --ai option adds semantic analysis using Claude Sonnet at about $0.012 per run.
What runtime or API keys are needed?
No configuration is needed for rules-based analysis. The --ai flag requires the ANTHROPIC_API_KEY environment variable. Sourcebook reads your repo’s git history and file structure directly.
Which languages are supported?
TypeScript/JavaScript, Python, Go, and Rust have full import graph and git analysis. Rust has partial convention detection. Other languages may work for co-change and test detection but without import graph support.
What is the license for Sourcebook?
Sourcebook is licensed under BSL-1.1 — source-available and free to use, but cannot be offered as a hosted service. It converts to MIT on 2030-03-25.
AI 与智能体 分类下的更多 MCP 服务器
Unreal Engine Generative AI Support Plugin
prajwalshettydevUnreal Engine plugin for LLM/GenAI models & MCP UE5 server. OpenAI GPT-5, Deepseek R1, Claude Opus/Sonnet, Gemini 3, Grok 4, Alibaba Qwen, Kimi, ElevenLabs TTS, Inworld, OpenRouter, Groq, GLM, Ollama, Local, Meshy, Tripo, Hunyuan3D, Rodin, fal, Dashscope, Seedream. NPC AI, agenti
LinkedIn MCP Server
stickerdanielOpen-source MCP server for LinkedIn. Give Claude and any MCP-compatible AI agent access to profiles, companies, jobs, and messages.
Shell and Coding agent for Claude and other mcp clients
rusiaamanShell and coding agent on mcp clients
🔎 GPT Researcher
assafelovicAn autonomous agent that conducts deep research on any data using any LLM providers
Model Context Protocol for Unreal Engine
chongdashuEnable AI assistant clients like Cursor, Windsurf and Claude Desktop to control Unreal Engine through natural language using the Model Context Protocol (MCP).
评论