Archon guide: setup and AI coding workflows

This live tutorial demonstrates Archon, an open-source harness builder for AI coding workflows. Cole Medin explains how workflows combine scripted steps with coding-agent tasks, then installs Archon on a Linux VPS, selects SQLite and configures authentication. He shows the CLI and web interface, workflow logs, isolated Git worktrees and human approval gates.

After a provider configuration problem interrupts the VPS demonstration, he switches to a working local instance. Four issue-fix workflows run concurrently and create pull requests. The session also covers artifact handoffs, fresh or continued context, and model selection in YAML. The demonstrated setup uses hosted Claude models, including Sonnet and Opus, through Claude authentication. It does not demonstrate offline inference, and hard token budgets are not available at the time of this stream.

The later demonstration builds a custom workflow inspired by GSD to research proposed Pi SDK support. Its YAML passes CLI validation, appears in the web graph and starts parallel research tasks. Approval steps expose missing artifact paths and answers that fail to reach later stages. The workflow needs further testing; the stream ends without a completed Pi integration or an autonomous deployment demonstration.