Hyperbeam documentation
Everything you need to run your team’s coding agents in the open: try it locally, learn how it fits together, and install it for your team.
Quick start
Run Hyperbeam on your laptop and connect your first coding agent.
Run it locallyConcepts
Workstreams, agents, sandboxes, harnesses and grants, and how they relate.
Learn the modelInstall on Kubernetes
Build the images, install the Helm chart, and point the harness profiles at your images.
Install for a teamGet started
Using Hyperbeam
- Start workDescribe an outcome and start an agent on it, from the board, a workstream or the command line.
- Watch and direct agentsFollow an agent as it works, reply to it, interrupt or stop it, and wake it up later.
- The board and workstreamsHow work moves from scoping to done, and what a workstream holds.
- ConductorsOne agent that scopes a request, starts workers across repositories, and is told when each reports back.
- NotificationsHear from your agents when they need you, in Hyperbeam and by direct message in Slack or Discord.
- Slack and DiscordStart agents from a channel, reply to them in the thread, and answer their questions without leaving chat.
- Skills and team notesWhat your team teaches its agents, kept in one place and shared by every session.
- WorkflowsSaved, parameterised ways of starting agents, written in YAML.
- Command lineStart, follow, message and stop agents from your terminal or CI.
Running a team
- Set up a teamFrom a fresh installation to a team that can start agents, in five steps.
- People and permissionsRoles, the grants that decide what each person's agents may do, and per-repository access.
- Harnesses and modelsWhich coding agents Hyperbeam runs, how to add your own, and how models are chosen.
- EnvironmentsGive agents your project's toolchains by running their sandboxes in an image of your own.
- CredentialsTeam and personal keys for harnesses, how they are stored, and what to know about subscription tokens.
- IntegrationsConnect GitHub or GitLab, first-class services like Linear, Notion, Slack, Datadog, Sentry and web search, and any other MCP server.
Deploy
- Run in a local clusterRun Hyperbeam on your own machine with the Helm chart, in a kind cluster, with every session in its own sandbox pod.
- Install on KubernetesBuild the images, install the Helm chart, and point the harness profiles at your images.
- Install on AWSRun Hyperbeam on one EC2 instance with the Terraform module, for about $115 a month.
- Install on a Raspberry PiRun Hyperbeam on a Raspberry Pi at home or in the office, with sandboxed agents, using k3s and one script.
- ConfigurationEvery server setting that matters, with its environment variable and default.
- Security modelWhat a sandbox holds, what it can reach, and how every action outside it is checked and recorded.