Favicon of Helix

Helix

Self-hosted AI agent orchestration on Mac, Linux or Kubernetes, with isolated desktops, human code review and local inference through Ollama or vLLM.

Screenshot of Helix website

Helix is a paid, self-hosted platform for engineering teams running coding agents in parallel. Each task gets an isolated Linux desktop with an editor, browser and terminal, so agents can build and test applications while teammates watch or take control through their browsers.

A shared Kanban board organizes work around specifications, with human review before implementation and pull request approval before merge. Chat history and running environments persist across team handoffs. Developers can inspect files, use the agent's terminals and pair program in the same workspace.

Helix works with Claude Code, OpenAI Codex, Gemini CLI, Qwen Code, Goose, Zed Agent and other ACP-compatible agents. You can change the agent or model during a task while keeping its conversation and uncommitted work. It connects to existing Claude or ChatGPT subscriptions, provider API keys and OpenAI-compatible endpoints.

The platform runs on macOS, Linux or Kubernetes. With local models through Ollama or vLLM, inference can stay on your hardware, including in air-gapped deployments. The Mac offering supports fully offline use with local models and at least 64 GB of unified memory. Credentials and development data stay on the host machine; choosing a cloud model provider sends inference requests to that service. Managed Helix Cloud hosts the agent infrastructure instead.

Team controls include role-based access, SSO, temporary task credentials, audit logs and usage tracking. Kodit provides code indexing and context across repositories, while the wider platform includes document retrieval, vision RAG and MCP tool support.

Similar to Helix