Red Queen – deterministic orchestration for AI coding agents

odyth1 pts0 comments

Red Queen — The Jenkins for Your AI Coding Workflow

Why Red Queen

You already have a coding agent. What you don't have is a deterministic pipeline<br>to shuttle tickets through it — spec, plan review, code, review, test, human gates —<br>while you're not looking. Red Queen is that pipeline. Configurable in YAML, auditable,<br>and built on a state machine that spends zero tokens on routing.

You pay for code, not coordination

The state machine is pure deterministic logic — no LLM in the routing layer. Every token you spend goes to actual code work, not deciding what to do next.

Isolated skill workers

Each phase — spec writer, coder, reviewer, tester, comment-handler — runs a purpose-built prompt in isolation. Focused beats kitchen-sink.

Human-in-the-loop gates

Review checkpoints are first-class workflow configuration, not an afterthought. Add them, remove them, reorder them in redqueen.yaml.

Issue tracker native

Bidirectional sync with Jira and GitHub Issues. Work flows from your tracker through the pipeline and back. Linear on the roadmap.

Retry with escalation

Failed phases retry up to three times, then escalate to a human gate. No infinite loops. No runaway token bills.

Webhook + polling

Optional webhook server for instant response, with a polling reconciler that works out of the box. No external queue required.

User-overridable skills

Skills are markdown prompt templates. Drop SKILL.md into .redqueen/skills// and your version wins over the built-in.

Adapter pattern

Every integration implements the IssueTracker or SourceControl interface. New trackers = new adapter, never a core change.

Works while you sleep

Background orchestrator keeps tickets moving through automated phases and parks work at human gates for when you're back at the keyboard.

How it works

You file a ticket. Claude writes a spec. An automated plan review scores<br>it for blockers and ambiguity. You approve at the human gate. Claude<br>writes the code and opens a PR. Another Claude reviews it. Another tests<br>it. You review the final PR and merge. Jump in at any gate, or rip the<br>gates out entirely.

spec-writing<br>plan-review<br>↻ spec-feedback · ≤3 retries<br>spec-review<br>coding<br>code-review<br>↻ coding · ≤3 retries<br>testing<br>human-review<br>merged

Automated (AI worker)<br>Human gate<br>Terminal

Phases are dynamic — defined in redqueen.yaml, not baked into the code.<br>Add a security-review gate. Remove the human spec-review and let<br>plan-review auto-promote clean specs with skipSpecReviewIfReady.<br>The graph is yours to shape.

What Red Queen is not

Red Queen sits above your coding agent — it doesn't replace it, host it,<br>or manage its skills. If you've seen one of these, Red Queen is a<br>different category of thing.

Not an agent host vs. opencode

Red Queen spawns claude CLI subprocesses and drives them through phases. It's the dispatcher, not the agent runtime.

Not a skill manager vs. agentskills.io

Our &ldquo;skills&rdquo; are phase prompt templates for the pipeline. They follow the agentskills.io format so they're portable, but Red Queen isn't a registry or loader for generic user skills.

Not a spec framework vs. OpenSpec

No persistent PRDs, no cross-session planning layer. Red Queen executes a ticket through a pipeline — complementary to spec frameworks, not a replacement.

Not an interactive coding tool vs. Claude Code, Aider

Red Queen runs in the background and moves tickets through phases while you're not looking. Use Claude Code or Aider for pair-programming at the keyboard.

Quickstart

From zero to your first AI-driven ticket in about 15 minutes.

Install Claude Code and run claude --version. Red Queen dispatches Claude Code workers, so the CLI must be logged in.

Install Red Queen:<br>npm install -g redqueen

Generate a GitHub PAT at github.com/settings/personal-access-tokens/new with Contents, Issues, Pull requests, Workflows, Metadata permissions on your target repo.

Initialize inside your repo:<br>redqueen init -y

Add your token to .env:<br>GITHUB_PAT=ghp_xxxxxxxxxxxxxxxxxxxx

Start the orchestrator:<br>redqueen start<br>Open http://127.0.0.1:4400 — the live dashboard.

Open a test issue in GitHub and apply the rq:phase:spec-writing label. Watch the phases move.

Full docs, adapter guides, and configuration reference live in the<br>GitHub README.

How Red Queen compares

Red Queen isn't a replacement for the tools below — it's the manager that delegates to them.<br>It dispatches Claude Code today; the adapter pattern means swapping in other coding CLIs is a config change, not a rewrite.

Red Queen<br>Claude Code (solo)<br>Aider<br>Cline / Roo<br>Devin / OpenHands

Form factor<br>Background orchestrator<br>Interactive CLI<br>Interactive pair<br>IDE agent<br>Autonomous agent

Orchestration<br>Deterministic state machine<br>You are the orchestrator<br>You are the orchestrator<br>IDE-driven loop<br>LLM-driven loop

Human gates<br>First-class, configurable<br>Ad-hoc<br>Ad-hoc<br>Ad-hoc<br>Minimal

Token cost for routing<br>Zero<br>N/A (you route)<br>N/A (you route)<br>Per decision<br>Per decision

Works while you...

queen review code claude spec human

Related Articles