AGENT RUNTIMES // YOUR CLUSTER
Run Codex, Hermes, and OpenClaw on Your Own Cluster
Deploy AI agent runtimes into your own Kubernetes cluster. Edka manages Codex environments, profiles, signed webhooks, state backups, logs, runtime health, and endpoints. Secrets and runtime state stay in your cluster.
FEATURES // AGENTS
Operate Agents Like Infrastructure
The Agents workspace deploys agent runtimes through the same pipeline as apps and exposes live runtime operations: profiles, webhooks, state backups, health, and logs. Runtimes run on your own cluster.
One-Click Deploy
Install agent runtimes from the catalog using the same pipeline as apps.
- Reliable deployment pipeline
- Required add-ons installed automatically
- Live deploy and readiness progress
Codex Environments
Repository-backed Codex workspaces you attach to from your local CLI.
- One isolated environment per repository
- ChatGPT or OpenAI Project API key auth
- Short-lived attach tokens after step-up
- Auto-delete TTLs and resource presets
Profiles
Run multiple agent identities inside one runtime pod.
- Named in-pod profiles
- Own instructions and memory
- Optional webhook routes per profile
- Start, stop, and restart per profile
Webhooks
Expose public, signed webhook routes for each profile.
- Public webhook routes per profile
- Signed payload verification
- Reveal the signing secret after step-up
- Send signed test payloads from Edka
State Backups
Export and restore agent state for recovery and cloning.
- Scheduled and on-demand backups
- Stored in your object storage integration
- Restore or clone profiles from a backup
- Credentials excluded (optional encrypted sidecar)
Health & Logs
Watch each runtime's health and endpoints in one place.
- Observed runtime health
- CPU, memory, and pod usage
- Runtime logs
- Dashboard, API, and webhook endpoints
Runs on Your Cluster
Agents run in your own infrastructure, under your control.
- Secrets and data stay in your cluster
- Cluster read/write access controls
- Device approval for browser pairing (OpenClaw)
- Archived clusters block mutating actions
GET STARTED // NOW
Deploy Your First Agent
Run Codex environments, plus Hermes and OpenClaw runtimes. Manage repository workspaces, profiles, webhooks, and state backups from the Agents workspace.
Where do AI agents run with Edka?
In your own Kubernetes cluster, on infrastructure you control. Secrets and agent data stay in the cluster, with cluster read/write access controls.
Which agent runtimes does Edka support?
Codex, Hermes, and OpenClaw. All install from the catalog through the same deployment pipeline as apps, with required add-ons installed automatically.
How do Codex environments work?
The Codex Agent creates isolated, repository-backed environments in your cluster, each with its own TLS endpoint. Sign in with your ChatGPT subscription or use a project-scoped OpenAI API key, then attach from the Codex CLI on your machine using a short-lived connection token. Provider credentials stay in Kubernetes Secrets in your cluster and are isolated from the workspace container.
Can one runtime host multiple agents?
Yes. A runtime pod hosts named profiles, each with its own instructions, memory, and optional webhook routes. Profiles start, stop, and restart individually.
Can agents receive webhooks from external services?
Yes. Each profile can expose public webhook routes with signed payload verification. You can reveal the signing secret after step-up authentication and send signed test payloads from Edka.
How is agent state backed up?
Scheduled and on-demand backups export agent state to your object storage integration. Credentials are excluded, with an optional encrypted sidecar. Restore or clone profiles from any backup.
What visibility do I get into a running agent?
Runtime health, CPU, memory, and pod usage, runtime logs, and the dashboard, API, and webhook endpoints, all from the Agents workspace.