// unrestricted ai agent
The Unrestricted AI agent that does the work
AiTM's autonomous agent brings unrestricted AI to your own environment. It plans, edits files, runs commands and browses the web to carry out authorized offensive-security work - and you stay in control through permission modes.
Unrestricted
Helps with in-scope work instead of refusing it
Your machine
Code and data never leave your environment
You decide
Manual, accept-edits or full auto
// install
Install in one command
Pick your OS - we detect it for you. Paste the command into your terminal to install or uninstall the desktop agent.
$curl -fsSL https://aitm.io/agent/install.sh | bashRuns in Terminal. Installs AiTM Agent.app to Applications and signs it so it launches cleanly.
// see it work
Unrestricted AI, with hands
The agent shows its work as polished activity cards - planning, searching, running tasks and editing files - never raw tool output.
Capabilities
Everything the engagement needs
One autonomous agent for authorized red teaming - from the first scan to the written report.
Local execution
Reads and edits files, runs shell and PowerShell, and manages background tasks - all inside the workspace you choose.
Web, browser & search
Searches and reads the live web and drives a browser to gather the context an engagement needs - CVE research included.
Recon to report
Plans the engagement, carries out in-scope recon, triages findings by severity and drafts the report for you.
Permission modes
Manual, accept-edits or full auto - enforced in the native core, so the agent never does more than you allow.
Nothing stored
The agent runs on your machine. The platform keeps no transcript and no source code - audit is metadata only.
Bring your own key
Point it at your own AI provider key. BYOK traffic is billed to your provider and never metered against your plan.
// in control
Autonomy on your terms
Set how much the agent does on its own. The permission mode is enforced in the native core - the interface can never override it.
Approve every action before it runs.
Auto-apply file edits, ask for the rest.
Let the agent plan and execute end to end.
Why an Unrestricted AI agent?
General agents refuse legitimate security work, hand back to-do lists instead of doing the work, and hide which model runs them. The AiTM agent is driven by the same Unrestricted AI that powers our assistant - tuned for authorized offensive security, so it helps with in-scope work instead of getting in your way.
It runs on your keys and your infrastructure, the model and prompts stay governed on the server, and nothing from your workspace is stored. Unrestricted where it counts, controlled where it matters.
- Built for authorized red teaming - not refusals
- Plans, executes and verifies, end to end
- Runs locally; your code never leaves the machine
- Model and provider never exposed
- Bring your own key - never metered against your plan
FAQ
Questions, answered
What is the unrestricted AI agent?+
A cross-platform desktop agent powered by AiTM's unrestricted AI. You describe the task and it plans it, reads and edits files, runs commands, browses the web and verifies its own work - autonomously, in your own environment.
What makes it “unrestricted”?+
It is built for authorized offensive security. Instead of refusing legitimate, in-scope work - recon, tooling, exploitation notes, CVE research and reporting - the unrestricted AI helps you get it done, within the scope you are authorized to operate in.
How much does the agent do on its own?+
As much as you allow. Choose manual (approve every action), accept-edits (auto-apply file edits and ask for the rest) or full auto (plan and execute end to end). The modes are enforced natively - the UI cannot bypass them.
Where do my code and data go?+
They stay on your machine. The agent executes locally and the platform stores no transcript or source code; for this surface the audit trail is metadata only.
Which AI model powers the agent?+
The underlying model and provider are never exposed - not in answers, headers, errors or the API model field. The model and the prompts stay governed on the server.
Is it only for authorized testing?+
Yes. The unrestricted AI agent is for engagements you are authorized to run, within your agreed scope and the law. Access is governed and abuse is not tolerated.
Authorized use & governance
The unrestricted AI agent is for security work you are authorized to perform, within your agreed scope and the law. Access is governed and abuse is not tolerated. You are solely responsible for your own actions - AiTM provides the platform and tooling.
Put the Unrestricted AI agent to work
Start free. Bring your own keys. Keep your code and data yours.
Join the community for updates, tips and support.
Join @aitmupdate