Welcome

Terragon is a developer platform that lets you run AI coding agents in the cloud. Whether you're at your browser, in your terminal, or on your phone, Terragon works wherever you are. Offload coding tasks and come back to pull requests ready to review.

With Terragon you get:

Isolated sandboxes: Every task gets its own virtual environment. Install deps, run tests/builds/lint, fetch web resources, while leaving your local setup untouched.

Parallel agents: Spin up multiple agents at once. Have one refactor auth, while another adds a feature and let them resolve merge conflicts.

Git-based workflows: Every task clones your repo into a new branch, commits as it works, and opens a pull request when work is complete.

On-the-go access: Start or manage tasks from a web dashboard, your phone, the terminal via the Terry CLI, GitHub comments and pull requests, or even from other tools you use via the Terry MCP server.

Secure environments: Code stays in your repos. GitHub & AI provider credentials are encrypted and used only inside your own environments, based on your actions.

Ways to Work with Terragon

• Use voice to describe tasks for Terragon while on your commute, then pickup where it left off when you're back at your desk
• Delegate small or medium tasks from your TODOs so you can focus on higher-priority work
• Give Terragon backlog tasks that are hard to prioritize relative to other work
• Tag @terragon-labs on GitHub issues to assign larger, well-scoped work to Terragon
• Terragon works well for exploring or prototyping ambiguous solutions iteratively. Give an ambiguous prompt to see how it approaches the problem, then use the Retry feature to refine the task

Types of Tasks That Work Well

• If you know how to do a task but want to delegate it, describe the task to Terragon as if you were instructing another engineer
• Tasks that replicate existing patterns in your codebase work well. If you have an existing feature to replicate or adapt, send that task to Terragon
• Give Terragon tasks that can be verified with high confidence (e.g., if CI passes)
• If the task seems too large for one person, it may be too large for Terragon. Consider breaking it into smaller tasks, or asking Terragon to plan its own work first

Get Started

Using Terragon

Resources