PentAGI

PentAGI is an open-source system of fully autonomous AI agents that perform complex penetration testing tasks. Instead of a single chatbot answering security questions, PentAGI deploys a coordinated team of specialized agents that plan an engagement, research targets, run real security tools in a sandbox, and report what they find.

Overview

PentAGI has gathered more than 31,600 stars on GitHub and remains under active development, with updates in October 2026. Built in Go with a React web interface and a GraphQL API, it runs self-hosted, typically via Docker. A primary agent orchestrates sub-agents for research, development and infrastructure work, delegating each step of the test. All tool execution happens inside isolated container environments stocked with a professional security toolkit, so autonomous actions stay contained. PentAGI works with OpenAI, Anthropic and local models, and keeps a searchable memory of past tasks so agents can learn from earlier engagements.

Key Features

• Multi-agent architecture with specialized researcher, developer and executor agents
• Fully autonomous penetration testing workflows, from recon to exploitation attempts
• Sandboxed Docker environment with a professional security toolset built in
• Support for OpenAI, Anthropic and self-hosted LLM providers
• Modern web UI plus GraphQL API for managing tasks and reviewing results
• Long-term memory and reporting so findings are documented, not just printed
• Self-hosted and open source under the MIT license for the core system

💳 Pricing

The PentAGI core is free and open source for self-hosting. Paid enterprise offerings and services are available from the vendor for teams that need support, advanced features or managed deployment.

Who Is It For

PentAGI is built for security engineers, red teams, penetration testers and bug bounty hunters who want to automate the repetitive parts of an assessment, and for researchers studying autonomous offensive-security agents.