Skip to content
Get packs

Guide · Open source

How to use PentAGI: setup, first run and tips

Updated · written with AI from the project's own documentation and checked against our catalogue data

PentAGI is an open-source, self-hosted system that takes a target and a goal described in plain language, plans the work with a team of AI agents, and runs security tools such as nmap, Metasploit and sqlmap inside an isolated Docker sandbox. This guide walks through what you need, how to get it running from the repository, and how to use it after login. It is written for people with some security and Docker background, and it assumes you only test systems you are authorized to assess.

PentAGIvxcontrol/pentagi · ★ 33k · MIT

Quick start

  1. 1Check prerequisites. Make sure Docker and Docker Compose are installed on the host you plan to use.
  2. 2Clone the repository. Run git clone https://github.com/vxcontrol/pentagi to download the project.
  3. 3Run the installer. Follow the Quick Start section in the README to run the interactive installer, which sets up PentAGI with Docker Compose.
  4. 4Configure an LLM provider. Pick a cloud provider, a custom OpenAI-compatible endpoint, or a local model through Ollama or vLLM, and follow its configuration section in the README.
  5. 5Log in to the web UI. Open the PentAGI interface and sign in to the system.
  6. 6Start a scoped test. Describe an authorized target and a goal in plain language and let the agents plan and run the tools.
  7. 7Review the report. Read the vulnerability report PentAGI generates at the end of the run.

What you need before you install PentAGI

PentAGI is self-hosted, so you run it on a server or machine you control. The README lists Docker and Docker Compose as the prerequisites. Plan for a Linux, macOS or Windows host that can run Docker containers.

  • Docker
  • Docker Compose
  • Git, to clone the repository
  • An LLM provider: a cloud API key, your own OpenAI-compatible endpoint, or a local model through Ollama or vLLM

Clone the PentAGI repository

The first installation step in the README is to clone the project from GitHub. The repository is vxcontrol/pentagi and it is published under the MIT licence.

git clone https://github.com/vxcontrol/pentagi

Run the interactive installer and start PentAGI

The README says PentAGI includes an interactive installer for Linux, macOS and Windows that sets the system up with Docker Compose. Follow the Quick Start section of the repository README for the installer steps, as they may change between releases.

PentAGI includes a web UI for system management and monitoring. After it is running, you log in to that interface to start working.

Choose an LLM provider

PentAGI does not force one model. The README lists support for more than 10 cloud LLM providers, plus aggregators, any OpenAI-compatible endpoint, and local models through Ollama or vLLM. Each provider has its own configuration section in the README, including Azure OpenAI, Ollama, OpenAI, Anthropic, Google AI (Gemini), AWS Bedrock, DeepSeek, GLM, Kimi, Qwen, MiniMax, Mistral and xAI (Grok).

If you want to keep data on your own hardware, the README points to Ollama and vLLM. For production local deployments it links to a vLLM plus Qwen3.5-27B-FP8 guide in the examples folder.

  • Cloud providers: OpenAI, Anthropic, Google AI/Gemini, AWS Bedrock and others
  • Local: Ollama or vLLM
  • Custom: any OpenAI-compatible endpoint

Use PentAGI after login

Once you are logged in, you describe a target and a goal in plain language. The agents then plan the steps, choose tools, run them in the Docker sandbox, store results and produce a report at the end.

Results and successful approaches are kept in PostgreSQL with the pgvector extension, so they can inform later runs. The README also describes an optional Neo4j knowledge graph built with Graphiti for tracking hosts, services and findings.

  • Describe the target and goal in plain language
  • Review the agents' plan and the tool runs as they happen
  • Read the generated vulnerability report, which includes exploitation guides

Use the REST or GraphQL API for automation

PentAGI offers REST and GraphQL APIs with Bearer token authentication. Teams can use them to automate testing and integrate the system with other tooling. The README includes an API Access section and an API Token Authentication feature for programmatic access.

Use PentAGI only for authorized testing

The project states that it is for authorized security testing only. Run it against infrastructure you own or have explicit written permission to assess. Keep the Docker sandbox in place, and note that PentAGI is a self-hosted tool, so you are responsible for securing the server it runs on.

What people use it for

  • Running an autonomous assistant on your own server to plan and execute nmap, Metasploit and sqlmap runs against systems you are authorized to test
  • Red team exercises against infrastructure you have explicit permission to assess
  • Testing with local models through Ollama or vLLM so data stays on your hardware
  • Automating repeated assessments through the REST or GraphQL APIs

Frequently asked questions

How do I install PentAGI?

Clone the repository from https://github.com/vxcontrol/pentagi, then follow the Quick Start section of the README. It describes an interactive installer for Linux, macOS and Windows that sets up PentAGI with Docker Compose.

What are the requirements for PentAGI?

The README lists Docker and Docker Compose as prerequisites. You also need an LLM provider, which can be a cloud API, an OpenAI-compatible endpoint, or a local model through Ollama or vLLM.

Is PentAGI free to use?

PentAGI is open source under the MIT licence, as stated in the repository. You still pay for any cloud LLM provider you connect, and you host the system yourself. Check the repository licence file for the full terms.

Can I run PentAGI with Ollama or local models?

Yes. The README lists Ollama and vLLM as supported options for local models, with a dedicated Ollama configuration section and a vLLM guide in the examples folder.

What are alternatives to PentAGI?

The comparable projects listed for this item are Nebula, PentestGPT, Strix and HackingBuddyGPT. Compare them on how they run, which models they support and how much setup they need.

Is PentAGI safe to use and is it legal?

The project states it is for authorized security testing only. Use it only on systems you own or have explicit permission to test, and keep its tools inside the isolated Docker sandbox.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

PentAGI suits security engineers, penetration testers and red teams who are comfortable running Docker on their own servers and who test only systems they are authorized to assess. If you want a fully hosted service with no server to manage, or you have no Docker or security background, pick a simpler or hosted tool instead.

Install options & downloads