AI Agents
A Lead Agent coordinates the work and knows when to stop.
Debatidor takes debate beyond a single answer. A Lead can turn a goal into a task tree, coordinate workers, and escalate critical decisions to the user.
What the Lead Agent does
Breaks down the goal
Creates subtasks with dependencies, state, and success criteria instead of issuing one monolithic instruction.
Assigns work
Routes every available task to the agent or worker with the right capabilities and context.
Detects blockers
Separates recoverable errors from actions that require permission, context, or a human decision.
Consolidates results
Combines progress, tests, and disagreements into a state the user can supervise and audit.
AI agents for remote development
A Debatidor worker can run in a repository on a VPS or another remote environment. The Lead dispatches a task, the worker reports logs and results, and the project can continue after the browser closes.
The goal is not unlimited autonomy. It is to separate routine work from irreversible decisions and keep a human at the right point in the loop.
Human approval at the risk boundary
When a task needs to overwrite sensitive configuration, rotate credentials, or perform another high-impact action, the Lead can pause and present the proposed change for approval. The user can approve, reject, or add instructions before execution continues.