Paseo + Kubernetes
A pod for every workspace.
Give each workspace its own Paseo daemon, CPU and memory limits, and persistent storage. Keep your main daemon connected to all of them.
Main Paseo daemonCompanion plugin + connected hosts
Workspace APaseo daemon · pod · PVC
Workspace BPaseo daemon · pod · PVC
Workspace CPaseo daemon · pod · PVC
More workspaces. Predictable resources.
Running many agents on one machine makes them compete for CPU, RAM and disk. Workspace pods put explicit limits around each environment, while an optional namespace quota caps aggregate admission.
2 CPUs4 GiB memory4 GiB ephemeral storage10 GiB persistent volume
Existing daemons, connected together
A Go operator reconciles a PaseoWorkspace into a pod, a Service and a persistent volume claim. The companion plugin on your main daemon registers ready environments as connected hosts.
- Agents execute inside the workspace pod. Their files, scripts and terminals stay with the workspace daemon.
- Communication uses Paseo’s host API. Discover a remote
serverId, create a workspace and start agents there. - Reverse registration is optional. Configure a reachable main-daemon connection so pod agents can initiate messages back.
- Storage survives a replacement pod. The replacement attaches the same PVC and retains the daemon’s local state.
Request. Connect. Work. Release.
kubernetes_workspace_create({ "name": "demo" })
kubernetes_workspace_list({})
# Wait for hostStatus: "online", then use the returned serverId.
# Export your results before deleting the environment.
kubernetes_workspace_release({ "name": "demo" })
The companion adds these tools to new agent sessions. Release deletes the environment’s pod and PVC; physical volume cleanup follows your StorageClass reclaim policy.
Initial implementation. Bring your own cluster and workspace image with agent providers. Remote Playbooks dispatch, idle shutdown and automatic repository checkout are not included yet.