Usage

Usage

Getting In

Open http://127.0.0.1:3000 (or your configured HTTPS origin). On a new installation, create the owner account. Registration closes after that first account; returning users log in.

After logging in, the Dashboard lists workspaces and recent sessions.

Starting a Session

Create or open a workspace, then start a session inside it. Sessions in one workspace share the same work host and folder.

Before your first session:

• Add and verify an orchestrator with Settings → Models → Test & Save.
• Save a local folder or tested SSH alias on the workspace's Connection page.

See Settings for details on each configuration panel.

The Chat Interface

The main workspace has two panels:

• Chat panel (left) - where you talk to the agent and see its output
• Shell panel (right) - live terminal sessions on the attack box

Sending Messages

Type your message in the input at the bottom of the chat panel. Describe what you want the agent to do:

• "Run a full port scan on 10.10.10.5"
• "This is a web application at http://target.htb. Find vulnerabilities and try to get a shell."
• "Analyze this request for injection points" (with a Burp request attached)

The agent will figure out which tools to use, run them, analyze results, and decide on next steps autonomously.

What You See

As the agent works, the chat shows:

• Reasoning - the agent's thought process (collapsible)
• Tool calls - each tool invocation with its arguments, live output, and exit status
• Analysis - the agent's interpretation of tool output and next steps
• Questions - if the agent needs information from you (target IP, scope, credentials), it will ask using the ask_user tool

Attaching Burp Requests

If you have Burp Suite configured, you can send captured HTTP requests to the workspace from the Burp proxy history page. These show up as an attachment chip in the chat input, giving the agent the full request context to work with.

Consent Model

The sidebar has a toggle that controls how the agent runs tools:

Mode	Behavior
Auto run	Routine actions run automatically. Built-in safety checks still stop high-impact operations for review.
Approve for me	The agent uses a separate safety review before automatically approving an action; uncertain or unsafe results fail closed.
Ask every time	Every tool action pauses for explicit approval.

Safety Protections

Even in auto-run mode, certain commands are flagged as dangerous and will always require explicit approval:

• Recursive forced deletion of system paths (rm -rf /)
• Writes to block devices (dd of=/dev/sda)
• Filesystem format commands (mkfs)
• Fork bombs
• System shutdown/reboot commands

The selected execution mode is stored with your user configuration and applies to main-agent, subagent, and swarm tool execution.

Manual Execution Fallback

If the selected work host becomes unavailable, the agent can switch to manual execution mode. Instead of pretending a failed command succeeded, it shows the command and waits for you to:

1. Copy the command
2. Run it on your attack box manually
3. Paste the output back into the provided text area

This keeps the engagement going even when connectivity is interrupted.

Shell Panel

The right side of the workspace shows live shell sessions on the attack box. The agent creates shells automatically when it needs them, but you can also:

• Spawn shells manually using the + button
• Switch between shells using tabs
• Type directly into any shell

Shells come in two types:

• PTY - interactive terminal (default for agent-spawned shells)
• Exec - non-interactive command execution

The shell panel also shows work-host connection status. SSH-backed workspaces can be re-tested from the Connection page.

Sidebar

The sidebar on the left gives you access to:

• Consent toggle - switch between "Ask consent" and "Auto run"
• Agent Tools - expand to see all available tools and toggle them on/off for the current session (see Agent Tools)
• Workspace navigation - links to the main workspace, VPN management, GUI desktop, and Burp proxy pages
• Clear context - wipe the conversation context (keeps the system prompt)
• Context usage - a visual indicator showing how much of the LLM context window is in use

Slash Commands

Type / in the chat input to see available commands. These are utility commands that run outside the normal agent loop:

Command	What it does
/help	List all slash commands
/status	Show current engagement status - targets, ports, vulns, credentials
/summarize	Summarize the entire session
/targets	Extract and list all targets/IPs mentioned
/export	Export findings as a structured report
/shells	List all active and closed shell sessions
/solve	Focus the agent on a synced CTF challenge (see CTF Solver)
/clear	Clear conversation context
/reset	Reset the session

See Slash Commands for details on each.

Tips

• Be specific about scope. Tell the agent exactly what's in scope, what isn't, and what kind of testing you want (black box, grey box, etc.).
• Use subagents. For broad engagements, the agent can spawn subagents to work on different tasks in parallel. You can also ask it to do this explicitly: "Run directory brute-force and subdomain enum in parallel."
• Check context usage. Long sessions accumulate context. If you're running low, use /summarize or /clear to free up space.
• Use the GUI. Some tools need a desktop (Burp, browser-based testing). The VNC desktop is accessible from the sidebar.