PentAGI

PentAGI (Penetration testing Artificial General Intelligence) is an open-source, self-hosted platform for automated security testing aimed at information security professionals, researchers, and ethical hackers doing authorized penetration testing. It runs fully autonomous AI agents that determine and execute pentesting steps inside a sandboxed, isolated Docker environment, with a built-in suite of 20+ professional tools such as nmap, metasploit, and sqlmap. Its architecture is a microservices system with a React/TypeScript web UI, a Go backend exposing REST and GraphQL APIs with Bearer-token auth, a PostgreSQL + pgvector vector store, and a multi-agent 'team of specialists' (orchestrator, researcher, developer, executor, plus adviser/planner/reflector supervision agents) that delegate tasks and collaborate. A smart memory system provides long-term, working, and episodic memory, augmented by a Graphiti-powered Neo4j knowledge graph for semantic context. Web intelligence comes from an isolated browser scraper plus integrations with search APIs including Tavily, Traversaal, Perplexity, DuckDuckGo, Google Custom Search, Sploitus, and Searxng. It supports 10+ LLM providers (OpenAI, Anthropic, Google Gemini, AWS Bedrock, Ollama, DeepSeek, GLM, Kimi, Qwen, plus OpenRouter/DeepInfra aggregators and custom OpenAI-compatible endpoints) and includes production-grade observability via Langfuse, Grafana/Prometheus, VictoriaMetrics, Jaeger, and Loki. Advanced agent supervision (execution monitoring and intelligent task planning) markedly improves results with smaller local models, enabling air-gapped autonomous pentesting. Deployment is via an interactive installer or Docker Compose (min 2 vCPU, 4GB RAM, 20GB disk). It generates detailed vulnerability reports with exploitation guides and web/Markdown/PDF export. Free and open source, fully self-hosted. Only test systems you are authorized to test.

What PentAGI does in practice

Fully autonomous AI pentesting in sandboxed, isolated Docker containers. Multi-agent team (orchestrator, researcher, developer, executor) with supervision agents. 20+ professional tools built in (nmap, metasploit, sqlmap, and more). Smart memory plus Graphiti/Neo4j knowledge graph and pgvector store. Supports 10+ LLM providers including local Ollama and vLLM for air-gapped use. Full observability (Langfuse, Grafana, Jaeger, Loki) and REST/GraphQL APIs. These capabilities should be judged as a workflow, not isolated checkboxes. Concrete best-fit scenarios include: Authorized autonomous penetration testing in self-hosted environments. Air-gapped or closed-network security testing with local LLMs. Security research and red-team automation with detailed reporting. Integrating automated pentesting into internal tooling via REST/GraphQL APIs.

Pricing and total cost

Open Source is listed at Free; Fully self-hosted via Docker Compose; Autonomous multi-agent pentesting with 20+ built-in tools; 10+ LLM providers plus custom endpoints; Knowledge graph, vector memory, and full observability stack; REST and GraphQL APIs with Bearer-token auth. These figures come from the staging record and vendor pages were fetched during this revision, but pricing is mutable. Recheck the live pricing page, checkout, or quote before publishing or buying. Include required seats, usage credits, model or API charges, storage, implementation, support, taxes, and overages. A free tier can prove compatibility; it does not necessarily model production volume or support needs.

Strengths and tradeoffs

Practical strengths include: Fully autonomous AI pentesting in sandboxed, isolated Docker containers. This directly supports the primary workflow. Multi-agent team (orchestrator, researcher, developer, executor) with supervision agents. 20+ professional tools built in (nmap, metasploit, sqlmap, and more). Smart memory plus Graphiti/Neo4j knowledge graph and pgvector store. Important constraints include: Plan prices can omit usage overages, model-provider charges, implementation work, taxes, or required seats. Features and limits can change by plan, billing interval, region, and release; verify current documentation. No native MCP support is documented in this record, so agent workflows may require API work or another integration layer. A representative pilot is necessary because a feature list does not establish reliability or operational fit. A capable product can still be the wrong operational choice. Document who reviews outputs, who owns failures, what data may enter the system, and how work continues during an outage.

How to evaluate PentAGI

Evaluate only systems you own or are explicitly authorized to test. Start with a disposable staging target containing seeded vulnerabilities and known clean controls. Track true positives, false positives, duplicate findings, time to validated exploit, remediation accuracy, and reviewer minutes per finding. Inspect every proposed patch, rerun regression tests, and verify scanners cannot reach out-of-scope assets. Require least-privilege credentials, immutable audit logs, retention controls, and a documented incident path. Autonomous execution increases the value of sandboxing and approval gates; a high finding count alone is not a useful success metric.

Bottom line

PentAGI is worth shortlisting when its capabilities match a measured workflow and the buyer can verify current limits. Do not select it from a feature list alone. Compare at least two alternatives using the same inputs, acceptance criteria, and cost model. Preserve test results so renewal is based on outcomes rather than adoption anecdotes.