Cortex
Analyzes a repository's history to generate structured .claude/docs/ context files for Claude Code, enabling informed code generation with historical insights, risk scores, and security notes.
README
Cortex
Project knowledge base for Claude Code.
Cortex analyzes your repository and writes .claude/docs/ — structured context files Claude Code reads before touching any file. It knows which parts of the codebase break most often, why certain decisions were made, and where the security risks are.
Not documentation you write. Knowledge extracted from your actual project history.
Install
curl -fsSL https://raw.githubusercontent.com/kovalenko-tech/cortex/main/install.sh | bash
Or with npx — no install required:
npx @kovalenko-tech/cortex-ai analyze
Requires Python 3.11+.
Quick start
cd your-project
cortex setup # interactive wizard
Or directly:
cortex analyze
git add .claude/
git commit -m "add cortex context"
What Claude Code sees
Before writing a single line of code, Claude Code gets this for every file:
> ⚡ Fresh — analyzed 2026-03-22 UTC
> 🔴 HIGH RISK (score: 65/100) — 12 bug fixes · no test coverage
## Historical Insights
- [Bug Fix] 2024-03-15: Session fails on UTC+0 timezone
Fix: Always use datetime.timezone.utc
## Decisions & Context
- ✅ Chose JWT over sessions for stateless auth — PR #234 @john
- 🚫 Don't call authenticate() before db.init() — PR #156 @maria
## Edit Checklist
- Run: pytest tests/test_auth.py -v
- Check: SESSION_TIMEOUT in config/settings.py
## Related Files
- tests/test_auth.py [co-change: 92%]
- middleware/session.py [co-change: 67%]
## Security Notes
- ✅ No issues found
Commands
cortex setup interactive setup wizard
cortex analyze analyze full project
cortex analyze --since HEAD~20 only changed files
cortex analyze --no-llm skip AI summaries
cortex analyze --no-cache force full re-analysis
cortex init generate CLAUDE.md
cortex context src/auth.py show context for one file
cortex status analysis status + freshness + risk
cortex risks files sorted by risk score
cortex freshness context staleness per file
cortex security security audit only
cortex deps dependency vulnerability scan
cortex mine-prs mine GitHub PR decisions
cortex diff main update changed files vs branch
cortex install-hook auto-update on every commit
cortex watch auto-update on new commits
cortex clean remove all context files
cortex mcp start MCP server for Claude Code
Claude Code integration
Add to .claude/settings.json for automatic context injection:
{
"mcpServers": {
"cortex": {
"command": "cortex",
"args": ["mcp"]
}
}
}
GitHub PR knowledge
Mine implicit decisions from PR discussions:
export GITHUB_TOKEN=ghp_...
cortex mine-prs
cortex analyze
AI summaries
export ANTHROPIC_API_KEY=sk-ant-...
cortex analyze
Uses Claude Haiku to generate a 2–3 sentence technical overview per file. Optional — all other features work without an API key.
Docs
Full documentation: https://kovalenko-tech.github.io/cortex/docs.html
License
MIT — kovalenko-tech/cortex
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。