RAPTOR is an autonomous security research framework that combines offensive and defensive capabilities. Built on Claude Code, it automates various tasks in security research, from scanning code using Semgrep and CodeQL to fuzzing binaries with American Fuzzy Lop (AFL). It not only identifies vulnerabilities but also generates proof-of-concepts for exploitation and offers patches to fix those vulnerabilities. Additional features include OSS forensics for GitHub repository investigations and a Cost Management system that tracks budgets and resource usage. The framework incorporates several new capabilities, such as a unified LLM interface for enhanced analysis and evidence collection from multiple sources like GitHub and the Wayback Machine. The architecture allows for parallel evidence gathering and sequential analysis, making it robust for forensic investigations. RAPTOR is open-source, which encourages community contributions and modular enhancements, although users are cautioned about its automatic installation of tools. Installation can be done directly or via a pre-configured devcontainer, which includes all necessary tools and dependencies. The devcontainer is substantial in size, around 6GB, and suited for complex analyses. Key features like Pydantic validation and smart model selection ensure users can navigate its functionalities efficiently. Pre-installed tools include Semgrep and CodeQL, while the offensive security agent specializes in tasks such as web application testing and binary exploitation.