A new entrant in the AI coding space, Agen wants to flip the developer workflow from "you drive, AI assists" to "AI drives, you review."
The pitch: spin up multiple AI coding agents that run simultaneously in cloud sandboxes. Each agent works on a task independently, fixes its own CI/CD pipeline errors, and delivers finished code you can inspect through a live preview. No local IDE required.
This positions Agen differently from tools like Cursor or Claude Code, which still operate as souped-up copilots inside your editor. Agen is closer to the "fleet of agents" model that Devin popularized - fully autonomous agents that take a task description and return a pull request.
The parallel execution angle is the interesting part. Most AI coding tools today run one conversation at a time. If you have ten bugs to fix, you fix them sequentially. Agen lets you dispatch all ten at once, each to its own sandboxed environment, and review the results as they come back.
The obvious questions: how good is the code? How well do agents handle ambiguous requirements? And what happens when Agent #3's changes conflict with Agent #7's? The platform is early, and these answers will determine whether parallel agents are genuinely faster or just parallel at producing code you still have to rewrite.
Pricing and model details weren't disclosed at launch. The product is available to try at agenhq.com.