Archon

Overview

Archon takes a different approach to AI-assisted coding than conversational agents like Claude Code or hermes-agent. Rather than reasoning interactively about what to do next, Archon encodes your development processes as YAML workflows — explicit DAGs that define exactly which steps run in which order, mixing deterministic operations (bash scripts, linters, tests) with AI-powered steps (code generation, code review). The result is an AI coding workflow that’s reproducible across projects and team members.

The analogy Cole Medin uses is apt: what Dockerfiles did for infrastructure (encode the build process in a portable, version-controlled format), Archon does for AI coding workflows. Your planning process, implementation approach, validation suite, and PR creation are all codified and version-controlled.

Key Capabilities

YAML DAG workflows: Define any multi-step development process as a directed acyclic graph. Each node is either a deterministic step (bash command, test run, linter) or an AI step (code generation, code review, summarization). The graph structure ensures dependencies are respected and outputs flow correctly between steps.

Git worktree isolation: Every Archon workflow run gets its own git worktree — changes don’t pollute your main working tree until you’re ready to merge. This makes it safe to run experimental or long-running AI workflows without destabilizing your development environment.

17 default workflows: Archon ships with pre-built workflows for common development tasks. You can also describe what you want in natural language and the router picks the appropriate workflow.

Script nodes: Inline TypeScript or Python scripts can be embedded directly in workflow definitions, with automatic dependency resolution via bun or uv runtimes.

Use Cases

Archon is most valuable for teams that want to standardize how AI assists with coding across projects. Instead of each developer using AI coding tools differently (and inconsistently), Archon encodes best practices in version-controlled YAML that runs the same way for everyone. It excels at workflows with clear validation criteria — where you can define “done” as a set of passing tests and linter checks, with AI steps generating the code that gets validated.

Considerations

Archon is a workflow engine, not a conversational agent. It doesn’t learn from sessions or build up memory of your preferences — each workflow run executes the defined DAG and exits. For exploratory or open-ended tasks, you’ll still want a conversational agent. Archon shines when the process is well-defined and repeatability matters more than flexibility.

Who It’s For

Engineering teams that want consistent, auditable AI coding workflows across projects. Particularly valuable for teams with established processes (specific review standards, required tests, PR templates) that can be encoded in YAML. Also useful for individuals who want to encode their personal AI coding workflow and run it reliably without configuration drift.