an agent operations layer for Claude Code & Codex
Six plugins that give Claude Code and Codex a lifecycle: plans live in your tracker, evidence lives in your repository, and every stage has to prove it happened before the next one starts.
Launching soon · 50 early-bird seats at $59 · no spam, unsubscribe anytime
Planning entropy
Specs, roadmaps and TODO files scatter across the repository. Each one was true the day it was written. Three weeks in, no one can say what is next, what is blocked, or which document still counts — so the plan quietly stops being the plan.
Delivery accountability
An agent closes a task and the diff looks fine. Was the test written before the fix? Did anyone review it who had not just written it? Was the checklist actually checked, or restated? Speed without evidence is just unreviewed change arriving faster.
LaunchOps fixes both with one rule: planning state lives in your tracker, technical evidence lives in your repository — and the agent is not allowed to skip either.
Each module owns one stretch of the road from idea to operating product. They hand off to each other in a fixed order, and a stage cannot be claimed without its evidence.
Turn a brief into initiatives, milestones and issues that carry acceptance criteria, roles and dependencies — not a markdown file nobody reopens.
Each issue moves through BDD and TDD stages that must record evidence before the next stage unlocks.
Route a change through accessibility, performance and security specialists that run locally against your repository.
Score launch checkpoints against recorded evidence and compute a go/no-go gate you can hand to a client.
Turn the post-launch plan into a channel strategy, a content calendar and a capacity plan you can hand to whoever executes them.
Layer positioning
Skill packs make an agent better at a task. They do not decide which task is next, who checks it, or whether it is finished. That is a different layer — and it is the one that breaks first once you have more than a weekend of work in flight.
Brainstorming, TDD, debugging, design review. Whatever your agent already knows how to do well.
Decides what runs when, who reviews it, what evidence is required, and what state the work is in.
Linear or GitHub Projects. Native issues, milestones and status — not a parallel universe of markdown.
Works with superpowers, not against it.LaunchOps runs alongside the skills you already have. It is not affiliated with, partnered with, or endorsed by any of them.
Six plugins, versioned and released together. Every version number below is the build shipped in the kit today.
RoleFlow for Linear — native initiatives, projects, milestones and issues with role phases and delivery gates.
The same RoleFlow lifecycle on GitHub Projects v2, milestones and issues.
Repo-level delivery lifecycle: BDD before TDD, evidence recorded at every stage, gates that refuse to skip.
Orchestrates installed specialist skills into one local audit run, preferring local source and local tools over hosted services.
Launch checkpoints with recorded evidence, a computed go/no-go gate, and sync to Linear or GitHub.
Builds the post-launch marketing plan: channels, calendar, capacity and the metrics that judge it.
Demo
This product is being launched with itself: the plan for the launch you are reading about lives in a real tracker, run by these plugins, with the evidence in a real repository. The walkthrough recording that shows it end to end is being cut now.
Walkthrough recording — in production
Get it in your inbox the day it goes live, along with the build-in-public receipts behind it.
Why this exists
Receipts from the build-in-public run — waitlist growth, early-bird seats sold, the readiness report for this very page — get published here as they happen, with the underlying artefact linked. Nothing goes on this section before it is true.
I kept writing specs and plans as markdown files next to the code. Every one of them was right when I wrote it, and useless a month later — once the product got big enough, I genuinely could not tell what to do next, what order things went in, or how anyone else would pick it up.
So I stopped keeping plans in files. Planning state goes in the tracker where it stays current; technical evidence stays in the repository where it belongs. That split is the whole product, and these six plugins are what enforce it.
No subscription, no seat metering, no licence server phoning home.
Early-bird checkout opens on launch day. Waitlist members get first access to the 50 seats.
Need more than one seat? Team license → contact
No. LaunchOps sits above them. Brainstorming still runs before planning, and TDD is still the inner loop of the build stage — LaunchOps decides what runs when, and records what happened. It is compatible with superpowers and runs alongside it; it is not affiliated with or endorsed by it.
Claude Code and Codex. The plugins ship both a .claude-plugin and a .codex-plugin manifest, and the governance rules are mirrored so a workflow behaves the same in either host.
You are invited as a read collaborator on the private launchops repository. Install with claude plugin marketplace add on the repo URL, then claude plugin install for each module. Nothing runs through a licence server.
Every future release of the six plugins in the kit, delivered to the same repository you already have access to. You run claude plugin update. New standalone add-ons built later are not automatically included.
Only the Plan module does, and it ships for both. The other four modules work on the repository alone — you can adopt them without a tracker and add Plan later.
Fourteen days, no questions asked, handled through Polar. Repository access is revoked automatically when the refund is processed.
No — the licence is per user. A team licence is planned; contact us and you will be told when it exists rather than sold something that does not.
Response within 48 hours on business days, through GitHub Discussions on the customer repository or support@launchops.dev. It is one person, and the SLA reflects that honestly.