What It Does
Dual Agent runs two AI instances that work together. You write an objective describing what you want built. One AI directs, the other builds. They keep going until your criteria are met.
You can watch the process, step in when you want, or leave it running and come back later.
How It Works
Write your objective
Create a simple document describing what you want: the goal, the tasks, and what success looks like.
Start Dual Agent
Point it at your objective file and a working directory. It sets up two separate AI sessions.
Watch or walk away
The proxy AI reads your objective and directs the worker AI. You can watch progress, intervene, or let it run.
Get results
The proxy verifies work against your criteria. When everything checks out, you have a working project.
Your Vision, Automated
The objective file is your document. Your technology choices. Your features. Your quality standards. Dual Agent executes your plan.
The Two Agents
The Proxy
- Reads your objective
- Creates a plan
- Directs the worker
- Verifies results independently
- Keeps pushing until done
The Worker
- Receives instructions
- Creates files and code
- Runs commands
- Reports progress
- Does the actual building
They run as separate sessions with independent context. The proxy doesn't just trust the worker's claims — it checks files, runs builds, and verifies tests directly.
Your Controls
- • Start — Begin autonomous execution
- • Intervene — Take over and give direct instructions
- • Pause — Save state and resume later
- • Verify — Run an independent check anytime
Good For
- • Building projects from clear specifications
- • Tasks that would take many back-and-forth prompts
- • Work you want to run in the background
- • Multiplying your output across multiple projects