owainlewis/machinist

What it solves

Machinist provides a self-hosted control plane for managing and orchestrating AI coding agents. It solves the problem of running autonomous coding tasks across your own infrastructure—such as local machines, VMs, or Kubernetes—while maintaining a centralized dashboard for tracking execution, approvals, and pull request delivery.

How it works

The project acts as a management layer that connects to your GitHub repositories. You define "agents" using configuration files, which then execute tasks using external coding CLI tools like Claude Code or Codex. The control plane tracks the entire lifecycle of a job—from the initial request or GitHub issue to the final pull request—and supports various workflows, including standard pipelines with multi-step reviews or single-agent automation for operational tasks.

Who it’s for

It is designed for developers and engineering teams who want to automate repository maintenance, issue triaging, or coding tasks using AI agents, but require full control over their infrastructure, data privacy, and execution environment without relying on third-party SaaS agent platforms.

Highlights

• Supports multiple execution targets: local host, remote VMs, or isolated Kubernetes jobs.
• Provides a React-based UI for managing runs, approvals, and repository configurations.
• Includes built-in support for scheduled automations and repository maintenance tasks.
• Offers flexible workflow modes, including multi-step pipelines with human-in-the-loop approvals or fully autonomous single-agent execution.
• Maintains a private runtime, ensuring that internal prompts, logs, and execution states remain under your control.