Environments Reusable, prebuilt cloud sandboxes for Factory Droids. An Environment packages the repositories, tools, and dependencies an agent needs so sandboxes are ready before work starts, supporting parallel missions, consistent automations, and team- or project-specific machine contexts.
Visit productFeatured media
Loading media…
View on XCosigns Received
“I've been moving more of my dev environment to cloud agents, especially for long-running @droid missions and loving the flexibility. Though reproducibility has been a pain @FactoryAI 's new environment blueprints tackle exactly that. Time to get droiding!”
“environments have been a game changer for me over the last couple weeks. I've been able to hillclimb continuously and run frequent automations without having to worry if my mac is awake and running. added bonuses are being able to check in on Droids from my phone and minimal set up thanks to one person creating the environment for all of eng”
“I’ve been using Environments to hill-climb on our evals, running experiments in parallel in ready-to-work sandboxes. That's resulted in fewer local worktrees and builds competing for my laptop’s resources, and more time comparing results and figuring out what actually improves our agents.”
“I have a cloud server that I use to run AI agents off my main computer (due to RAM conflicts and Windows issues). This new release is going to save me ~$65 a month on that server. It’s amazing how technology just keeps advancing! Congrats to @FactoryAI team!”
“i was recently reflecting on what features we've shipped in the past 3 months that feel like step function increases in the team's productivity. sandboxes are number one on the list. being able to reliably scale the number of running agents into the thousands is a massive force multiplier. for example, just in the last hour i was able to kick off parallel tasks to: - hillclimb on agent harness improvements - redesign our cloud persistance layer - implement several features for #2 on this list of most impactful products 👀 - write new lint rules for our monorepo there has been some discourse online about slop and how to avoid it. agents are strong enough nowadays that you need be parallelizing them, and sandboxes are incredibly important component used to fight the slop. using them, you can ensure every agent is equipped with the toolchain, skills, credentials, etc needed to be maximally productive in your project. you can also literally engineer your codebase into a slop-fighting machine (as @poteto rightly calls out) by adopting stronger type systems and more static analysis.”
“This is the first time I have found sandboxes to be both reliable and easy to configure. Parallelize away with Environments.”
“Droid computers are great. On-demand zero-setup Environments are better. No more thinking about OOM from too many concurrent sessions doing worktree checks on one machine. Effortless, even accidental scaling as I engage with threads. 🪄”
“Got early access to Factory Environments and it clicked right away define the sandbox once (repos + setup scripts), Factory prebuilds it and it’s always ready for work, from anywhere Nice one @FactoryAI 🔥”
“Have had the opportunity to do some early testing with Environments & I’m excited. Native sandboxes with lightening quick deployments for your droids. I love when teams cook on features 😮💨”