Introducing OpenCompute: a self-hosted orchestrator for agent sandboxes on your choice of runtime. Adopting autonomous software development should not require handing over your valuable code and context to someone else’s cloud. Scale your fleet of remote agents while keeping them in your VPC.

Meet the demand for cloud agents running across entire organizations. Enable interactive sessions as well as agents operating independently as work arises. Deploy OpenCompute once into your environment, and let it handle orchestration across your entire fleet of agent sandboxes.

Scalable cloud agents, securely in your environment, with zero infrastructure hassle. OpenCompute handles the sandbox lifecycle for you: spinning new machines up as they are needed, and down once sessions are over.

Organizations shouldn't need separate infrastructure for every kind of agentic workload. With OpenCompute, interactive sessions, scheduled Automations, and incident investigations all request the environments they need through the same lifecycle layer, whether Factory or the customer operates it. Agentic workloads will continue to run across many machines and clouds, and OpenCompute is built to orchestrate sandboxes across them.

OpenCompute is now available in private preview. Read more about OpenCompute:

Damn, so many new drops.

Everyday new updates, team is working hard

NGL This is awesome! opencompute for agents this crazy

Thank you!

Congrats on the OpenCompute launch, a self-hosted orchestrator for agent sandboxes is a great addition. Thanks for sharing this.

damn factory ai is shipping like crazy hard to catch up; time to lockin

Shipping every day!

You can only get this from Factory right now

self hosted sandboxes is a nice option for teams that cant hand their code to another cloud

This is the first design I’ve seen that actually treats the compute plane as customer-owned infrastructure instead of a thin veneer over the vendor’s cloud. Outbound WebSocket + durable desired state in the customer’s Postgres + gRPC drivers is the right separation. The interesting part will be how well the optional profiles (suspend/snapshot/fork) hold up once people start using mixed backends in the same fleet. Curious how you’re handling driver versioning and rollback when a provider changes its API under a live reconciler.

Ok I see you (again) @FactoryAI 👀

Do I need active subscription to use this, as I need for BYOM(achine)?

👏👏👏

Sandboxing is well understood; state is where these systems break. Keep the workspace and logs outside the sandbox so any node can resume a run after a restart. Game servers on Flux work the same way: the world outlives the container.

Self-hosted agent sandboxes make control a product feature, not just an ops choice, when code and context must stay inside the team’s boundary.

👏🙌

the self hosting argument is stronger than the post makes it. went through one of the popular free gateway plugins line by line last week and it ships a literal "bearer public" header to a third party while the model cards say prompts may be used to improve their services. people hand over code and context without knowing a gateway is in the path at all.

keeping the code in your own vpc is the decision most teams skipped when they signed up for a cloud agent

Your runtime, your sandbox, your rules. Ship it.

was just looking for something like this!!

Awesome !!!

Self-hosting addresses data residency, but production rollouts also need policy boundaries per agent: short-lived credentials, network egress controls, and clean sandbox teardown. Are these enforced by the orchestrator or delegated to each runtime? #AIAgents #Security

No thanks, I prefer Devin.

Self hosted agent orchestration in your own VPC finally scaling without giving up your code Huge for autonomous dev

Interesting idea... When are you open sourcing this?

@grok please explain this to me.

OpenCompute's docs separate sandbox placement, inference and session storage, which is the distinction I'd keep front and centre. To me, "in your VPC" only answers one part of where the code and context can go.

Keeping the sandbox in your environment is half the win. The other half is proving what the agent did inside it without shipping that context to a third party.

Keeping the code on your own machines is going to matter more than people think.

Love this approach self hosted and in your own VPC is exactly how agent fleets should scale securely!

Are credential revocation on teardown and fresh issuance after restore covered by the driver conformance tests, or handled by each deployment? I'd want that contract explicit before enabling snapshots for agents with write access.

Factory's OpenCompute empowers secure self-hosted development, keeping valuable code private while enabling autonomous innovation brilliantly

Is this open source or not?

Waiting your dm

Solid approach to private agent orchestration

Will be super helpful to orgs!