Roadmap
KubeOpenCode is in early alpha (v0.1.x). The API (v1alpha1) may introduce breaking changes between releases. We do not guarantee backward compatibility at this stage. See Getting Started for details.
Direction 1: Instant Messaging Integration
Goal: Make AI agents part of your team's daily communication workflow.
- Slack integration — Bi-directional: receive commands from Slack channels, push task results and notifications back
- Other IM platforms — Microsoft Teams, Lark/Feishu, and other enterprise messaging tools
- ChatOps patterns — Trigger tasks, check agent status, and review results without leaving your messaging app
Plugin infrastructure (ADR 0034 Part 1) is implemented. Slack integration (Part 2) is pending.
This direction focuses on usability — reducing the friction between "I need an AI agent to do something" and getting the result.
Direction 2: Kubernetes Ecosystem Integration
Goal: Production-grade stability and security through deeper integration with the Kubernetes ecosystem.
- GitOps — Native integration with ArgoCD, Flux for declarative agent management
- Policy engines — OPA/Gatekeeper, Kyverno integration for agent governance
- Network security — NetworkPolicy templates, service mesh integration
- Multi-tenancy — Namespace-level isolation, resource quotas, priority classes
This direction focuses on stability and security — making KubeOpenCode ready for production environments.
Direction 3: Observability (OpenTelemetry)
Goal: Full observability for AI agent workloads — understand what agents are doing, how they perform, and where they fail.
- OpenTelemetry integration — Instrument controller and agent lifecycle with OTel traces, metrics, and logs
- Task execution tracing — End-to-end distributed traces from Task creation through agent execution to completion
- Prometheus metrics — Task duration, success/failure rates, queue depth, agent utilization, token usage
- Structured logging — Consistent, queryable log format across controller, init containers, and agents
- Dashboard templates — Pre-built Grafana dashboards for task throughput, agent health, and error analysis
This direction focuses on operational insight — giving platform teams the data they need to run AI agents reliably at scale.
Direction 4: Container-in-Container (DinD) Support
Goal: Enable agent workloads that require running Docker, Podman, or Kind clusters inside containers — without compromising cluster security.
AI agents often need to run E2E tests against Kind clusters, build container images, or operate full development environments. These workloads require nested container runtimes inside the agent pod, which standard Kubernetes does not support out of the box.
KubeOpenCode itself does not need code changes — the Agent podSpec already supports runtimeClassName passthrough. What's needed is cluster configuration guidance: documentation and best practices for administrators to enable DinD capabilities using approaches like Sysbox, Rootless Podman, or Kata Containers, depending on their security and platform requirements.
See ADR 0029 for the full evaluation of approaches and phased adoption strategy.
Recently Added
These features have been implemented and are available in the current release:
| Feature | Description | Docs |
|---|---|---|
| Plugin Support | First-class OpenCode plugin support with npm-based installation | Plugins |
| Task Session Integration | Session info, token usage, and cost in Task status | Task Session |
| Task Timeout | Automatic timeout for long-running tasks | Task Timeout |
| Task Stop | Stop running tasks via annotation | Task Stop |
| Agent Share Link | Share terminal access via URL | Share Link |
| Git Auto-Sync | Automatic sync with remote Git repositories | Git Auto-Sync |
| CronTask | Scheduled and recurring task execution | CronTask |
| Standby Mode | Automatic suspend/resume for idle agents | Persistence |
| Extra Env & System Containers | Additional environment variables and sidecar containers | Pod Configuration |
Direction 5: Agent Registry (In Progress)
Goal: Enterprise agent catalog and visual agent assembly marketplace.
- Visual Agent Builder — UI for assembling agents from templates, skills, and plugins
- Public Registry — Community-driven catalog of AgentTemplates, skills, and plugins
- Private Registry — Enterprise registry for internal team templates and credentials
See ADR 0036 for the design.
Deferred
Token Usage Tracking & Cost Reporting
Status: Waiting for upstream OpenCode support
Track per-Task token consumption and estimated cost. Blocked on OpenCode lacking machine-readable output for token statistics. See ADR 0013 for details.