Skip to content
Hyperfluid 2.0 is live console.hyperfluid.cloud/signup
AI Dev Workstations

Sovereign code agents, sandboxed in your cloud

One click, one repo, an agent session backed by your in-cluster GPUs

Each workstation is an isolated sandbox, ready in minutes: the open source code agent clones your repository and works with the models served inside your cluster. A browser session with terminal and Git stays available to take over.

Hyperfluid Dev Workstations: sandboxed code agents in the console

Specifications

Code agent
Open source tool, one session per repo
Isolation
Isolated sandbox
Inference
Models served on your cluster GPUs
Startup
One click, an isolated sandbox

Use cases

Launch an agent on a repository

One click, one repo: the workstation clones your Git repository and opens an open source code agent session, isolated from the rest

Disposable, sandboxed workstation

Keep code and AI inside your cloud

Each workstation is an isolated sandbox. Point the agent at the models served inside your cluster and both code and inference stay with you

Inference on your GPUs

Take over whenever you need

Workstation templates to standardize environments, and a browser session with terminal and Git to inspect, fix and validate the agent's work

Terminal and Git included

In action

Hand a task to a code agent

You delegate a change to a code agent without your repository leaving your infrastructure

  1. Create a workstation from the console: Git repository, branch and repository access secret
  2. Pick the environment from the template catalog, then the inference model served inside your cluster
  3. The workstation starts in an isolated sandbox and clones the repository
  4. Describe the task to the agent in the Agents tab, then follow its work in the terminal and the editor

A task delegated to an agent, with code and inference kept inside your cluster

Standardize the team's environments

A platform team wants every developer to start from the same tooling, reproducible on every restart

  1. Publish a workstation template with the expected base image and tooling
  2. Each developer selects that template at the Environment step of the creation flow
  3. The tooling is declared in the template, so it is reinstalled identically on every provisioning
  4. Review the changes in the Source Control tab before pushing the branch

Reproducible environments for developers and agents alike

Key benefits

  • Sandboxed code agents: one open source agent session per repository, connectable to the models served on your GPUs
  • Isolated sandbox for every workstation, with its terminal and its coding agent
  • Inference on the models served inside your cluster: code and AI requests stay with you
  • Workstation templates and a browser session, with terminal and Git, to stay in control

Ready to run code agents in your cloud?

Request a demo of Dev Workstations.