S1N6H/pentest-harness

What it solves

Pentest Harness provides a professional, dedicated workspace for security researchers and hackers. It streamlines authorized penetration testing, bug bounty research, and CTF engagements by integrating AI agents capable of executing technical tasks directly within a secure, dark-themed interface.

How it works

The project is built on a plugin architecture (Cordis) where model adapters, tools, and sessions are replaceable. It connects to various AI model APIs (OpenAI, Anthropic, DeepSeek, Google, etc.) via a multi-provider LLM engine. The agent has access to a full toolset including a shell, filesystem access, web research capabilities, and the ability to manage subagents and background jobs.

Who it’s for

Security professionals, bug bounty hunters, and CTF participants who want to leverage AI agents to automate and orchestrate security testing workflows.

Highlights

• Multi-provider LLM support: One-click auto-discovery for models from major providers and OpenAI-compatible gateways.
• Durable sessions: Uses JSONL/SQLite persistence to allow users to resume engagements exactly where they left off.
• Secure credential storage: API keys are stored in an owner-only credential store rather than in settings files or logs.
• Context management: Features token metering, automatic compaction, and tool-result pruning to maintain long-term context.
• Extensible design: Everything is implemented as a plugin, ensuring no closed core components.