Run AI agents and coding agents on your cluster. Model keys never leave it.
Describe work as a Task. Orka runs it in a Pod, keeps a durable record of what happened, and hands you the result through an API, a CLI, or the built-in dashboard. One Helm command to install.
Quick start
Install Orka on Kubernetes and run your first task.
1. Install with Helm
One Helm command. The chart creates its namespaces, its encryption key, and its webhook certificate. Works on kind or minikube. No model API key needed yet.
2. Run your first AI task
Connect to the API, add your Anthropic, OpenAI, or Azure OpenAI key as a Provider, and submit a Task.
For development, build from source.
Highlights
AI Agents
Anthropic, OpenAI, or Azure OpenAI agents with tools, skills, and session persistence.
Multi-Agent Coordination
Coordinators decompose tasks and delegate to specialists with depth and concurrency controls.
Interactive Chat
An agentic orchestrator with SSE streaming that creates and manages agents and tasks for you.
Durable Memory
Namespace-scoped recall, transcript search, and reviewable memory proposals for coordinated agents.
Repository Security Scanning
Scheduled scans with threat models, validated findings, patch generation, and remediation PRs.
OpenAI & Anthropic APIs
Drop-in compatible endpoints for Continue, Cursor, Claude Code, and any OpenAI-native client.
Providers & Runtimes
Bring your own models and coding agents — Orka keeps the credentials in the cluster and gives them a common API.
Anthropic
Claude models through the Provider CRD and an Anthropic-compatible Messages API.
OpenAI
GPT models behind an OpenAI-compatible chat completions endpoint.
Azure OpenAI
Enterprise OpenAI deployments running on Azure.
ACP RuntimePools
Run Codex, OpenCode, Claude, and Copilot in digest-pinned, scale-to-zero RuntimePools with brokered tools and clean-room delivery.
Join the community
Issues, ideas, and pull requests are all welcome.