The Platform for Remote
AI Coding Agents

Deploy AI agents that write code, run tests, and push to GitHub — on their own machines, around the clock. Use Codex, Claude, or bring your own model. Watch everything in real time.

Deploy an agentCompare tools

What Is a Remote AI Coding Agent?

A remote AI coding agent is an AI-powered developer that runs on its own machine in the cloud. Unlike editor-based AI tools like Copilot or Cursor, remote agents work independently — they don't need your editor open, your laptop awake, or you watching.

You describe a task in plain language. The agent writes code, installs dependencies, runs tests, fixes failures, and pushes the result to GitHub. You can watch the entire process through a live terminal, or check back later when it's done.

chat.dev supports two leading AI models as remote agents: remote Codex agents (OpenAI) and remote Claude agents (Anthropic). Each agent gets its own isolated machine with a persistent filesystem — your work is always there when you come back.

The difference between a copilot and a remote agent is the difference between pair programming and delegation. A copilot helps you write code faster. A remote agent takes ownership of a task and delivers finished work.

This changes the economics of software development. A remote codex agent working overnight costs less than $1 in compute. A junior developer doing the same work costs $200-400. Remote AI agents aren't replacing developers — they're multiplying what each developer can accomplish.

On chat.dev, you can run multiple agents in parallel. One agent refactors your auth module while another writes tests for billing while a third prototypes a new feature. You review three PRs in the morning instead of writing three features yourself.

AI Coding Agent Comparison 2026

How does chat.dev compare to Devin, Cursor, GitHub Copilot, and Replit Agent? Feature-by-feature comparison.

Featurechat.devDevinCursorCopilotReplit
Runs autonomously (no IDE open)
Own machine per agent
Live terminal you can watch
Works while laptop is closed
GitHub auto-commit and push
File browser with drag-drop
Voice input
GPU machines available
Choose your AI model
Bring your own API key
Starts at under $30/mo

Last updated April 2026. Based on publicly available features.

What Can Remote AI Agents Build?

Full-stack applications

"Build a Next.js app with Stripe billing, email auth, and a dashboard." The agent scaffolds the entire project, wires up the API, writes tests, and commits.

Bug fixing

"This test is failing on CI. Here's the error log." The agent traces the bug, writes a fix, verifies all tests pass, and opens a PR.

Test suites

"Write integration tests for the payments module. Cover the happy path, error cases, and webhook handling." Wake up to 40 new tests.

Refactoring

"Migrate from Express to Hono. Update all route handlers, middleware, and tests." The agent handles the tedious work across hundreds of files.

DevOps & CI/CD

"Set up GitHub Actions for lint, test, build, and deploy to Fly.io. Add a staging environment." Infrastructure as a task.

Documentation

"Write API docs for every endpoint. Use JSDoc format. Generate an OpenAPI spec." The agent reads the code and documents what it finds.

How Remote AI Agents Work on chat.dev

1

Create

Pick a name, choose your AI model (Codex or Claude), select a machine tier, and optionally connect a GitHub repo.

2

Instruct

Describe your task in natural language via chat, terminal, or voice. Be specific — the agent follows your instructions literally.

3

Watch or wait

Monitor the agent through a live terminal in your browser, or close your laptop and let it run. It works independently.

4

Review & ship

Check the code in the file browser, review the git diff, merge the PR. Or let the agent push directly to your branch.

Pricing for Remote AI Agent Hosting

Compute hours and API credits included. No credit card required to start.

Base

$29.99/mo

10,000 compute hours
$10 OpenAI credits
All features

Pro

$99.99/mo

20,000 compute hours
$30 OpenAI credits
All features

Expert

$499.99/mo

50,000 compute hours
$100 OpenAI credits
20% off upgraded machines

Machine tiers: Standard ($7/mo) | Pro ($39/mo) | Max ($130/mo) | GPU ($3,432/mo)

Frequently Asked Questions

About remote AI coding agents, pricing, security, and how chat.dev works.

What is a remote codex agent?
A remote codex agent is an AI coding agent powered by OpenAI's Codex model that runs on its own machine in the cloud. You give it tasks through a chat interface or live terminal, and it writes code, runs tests, and pushes to GitHub autonomously. It keeps working even when your browser is closed.
What is a remote claude agent?
A remote claude agent is an AI coding agent powered by Anthropic's Claude model that runs on its own machine in the cloud. It has the same capabilities as a remote codex agent — full development environment, GitHub integration, persistent workspace — but uses Claude's reasoning abilities.
How is chat.dev different from GitHub Copilot?
Copilot works inside your editor and suggests code as you type. chat.dev agents work independently on their own machines. You describe a task, the agent builds it end-to-end, and you review the result. Copilot is a pair programmer; chat.dev agents are autonomous developers.
How is chat.dev different from Devin?
Both chat.dev and Devin are autonomous AI coding agents. The key difference is transparency: chat.dev gives you a live terminal where you can watch every command and interact with the agent in real time. chat.dev also lets you choose your AI model and machine tier.
Can I run Codex remotely in the cloud?
Yes. chat.dev lets you deploy a remote codex agent in the cloud in about 30 seconds. The agent gets its own machine, connects to your GitHub repos, and works autonomously. No local setup required.
Can I run Claude remotely in the cloud?
Yes. chat.dev supports Claude as a remote AI coding agent. Same capabilities as remote codex agents: own machine, live terminal, GitHub integration, persistent workspace.
Is my code safe?
Each agent runs on its own isolated machine. Your code and credentials are not shared between agents or users. GitHub tokens are scoped to the repos you authorize.
What languages do AI coding agents support?
Remote codex and claude agents can write code in any language. The machine comes with Node.js, Python, and Git. The agent can install any additional tools it needs.
Can I run multiple agents at once?
Yes. You can deploy as many agents as your subscription allows and run them in parallel. One agent refactors auth while another writes tests while a third builds a new feature.
How much does it cost?
Subscriptions start at $29.99/month with 10,000 compute hours and $10 in OpenAI credits. Machine tiers range from $7/month (standard) to $3,432/month (GPU). No credit card required to sign up.
What if the agent makes a mistake?
You can watch the agent work in real time and interrupt it at any point. The agent also commits its work to git, so you can always revert. And because it runs on its own machine, mistakes don't affect your local environment.
Can I use my own OpenAI API key?
Yes. You can bring your own API key or use the credits included with your subscription.

Company

chat.dev is built by Antipodal, Inc., a company focused on making AI coding agents practical and accessible.

We believe the future of software development is AI agents that work independently — not as assistants in your editor, but as autonomous developers with their own machines and their own workflows. chat.dev is the infrastructure that makes this possible.

Deploy a remote AI agent in 30 seconds.

No credit card. No setup. Sign in and go.

Get started