Introducing PentAGI Enterprise 3.0

PentAGI Enterprise 3.0 is for teams that deliver penetration tests: consultancies, in-house red teams and MSSPs. It shares the self-hosted platform of the open-source edition — flows, sandbox, tools and memory — and adds what turns an autonomous run into a deliverable.

Reports your client can read

Reports are built from templates you manage in Settings → Report Templates. A composer agent writes the narrative sections from the flow's record, while the cover, the metrics and tables such as the severity distribution, the step-by-step attack chain and the tools used are assembled without a model. Repeating sections expand once per item — one chapter per finding, each with a severity rating, proof of concept, impact and remediation steps.

Before delivery, diagnostics check the draft without a model: they flag identifiers that appear nowhere in the flow and a report started while work was still running. They advise and never rewrite. Ask for corrections in plain language and a reviser agent applies them. Export to PDF or DOCX with your logo and cover page, or to Markdown.

Agent profiles and sub-agents

An agent is defined by a profile: role, base prompt, model preset and tools. Built-in profiles cover working agents such as pentester, coder, researcher and recon, the interactive assistant, and the pipeline agents that plan and report. A planner picks the profile for each subtask, and you can add your own in Settings → Agent Profiles.

Working agents can start sub-agents for parallel work inside a subtask. Time budgets handed down from the task to each subtask, a cap on subtasks per task and a ceiling on concurrent sub-agents keep that delegation under control.

Plugins and your own MCP servers

Agent Plugins package skills, MCP server declarations and agent profiles in the open agent-plugins.org format. Planners see a short description and the skill names of each plugin; the skills of the ones they choose go into the working agent's prompt.

Connect your own MCP servers over Streamable HTTP or HTTP with SSE. Credentials are supplied when a server is installed and never returned by the API.

The attack chain, live

The knowledge graph links hosts, ports, services, vulnerabilities, confirmed access and accounts as the agents work. With the knowledge graph enabled, the flow's Dashboard → Analytics tab shows it as the main attack chain, the action chain, the infrastructure, the access chain and the privilege escalation chain, with a timeline to replay how the picture developed and tables of open ports, detected CVEs and credential status.

Your terminal next to the agents

Open your own terminals in the flow's sandbox and work side by side with the automation.

Get Enterprise

Licenses are issued and managed in the PentAGI Console. See the edition comparison and the Enterprise documentation.