Introduction

Agent HQ is an AI-powered development workspace. You describe what you want built, and autonomous AI agents write the code, run the tests, and submit pull requests.

How it works

Create a project linked to a GitHub repository
Chat with Pilot to describe what you need
Pilot creates tasks and executes them autonomously
Review the results — code diffs, test results, and live previews
Approve and merge the generated pull request

Core principles

Conversation-driven development

Instead of writing tickets and specifications, you have a conversation with Pilot. It asks clarifying questions, breaks work into tasks, and executes them. You can guide the process at any point — approve, redirect, or take over manually.

Autonomous execution

Tasks run in isolated sandboxes. Each task gets its own branch, cloned repo, and execution environment. Pilot can write code, install dependencies, run tests, take screenshots, and create pull requests — all without manual intervention.

Human in the loop

You stay in control. Dangerous operations require approval. You can interrupt any running task. Every change is on a separate branch so nothing touches your main codebase until you approve it.

Multi-model flexibility

Choose the AI model that fits each task. Use Claude for complex reasoning, GPT-4 for creative writing, Gemini for fast iteration, or Cloudflare’s free models for lightweight tasks.

Who is it for?

Solo developers who want to move faster by offloading routine tasks
Small teams who want to scale their output without scaling headcount
Agencies who need to prototype and iterate quickly for clients
Anyone who wants an AI teammate that actually writes, tests, and ships code

What can Pilot do?

Write new features from a description
Fix bugs given a reproduction or error message
Refactor and improve existing code
Write and run tests
Create pull requests with clear descriptions
Take and annotate browser screenshots
Execute multi-step workflows that span multiple files