Aakash GuptaBuild a Growth Operating System in Claude Code, Step by Step
At a glance
WHAT IT’S REALLY ABOUT
An agentic growth OS: goals, skills, and tool-connected Claude workflows
- James McAulay presents “Growth Agents,” a GitHub-based agentic growth operating system that turns Claude Code into a structured environment for goals, routines, and repeatable growth skills.
- The core workflow is a goal cascade (mission → quarterly OKRs → monthly/weekly retros → daily standup) designed to keep AI work aligned and prevent rabbit holes.
- He demonstrates practical growth skills—especially an SEO audit pipeline using Firecrawl, DataForSEO, and Lighthouse—to generate actionable reports and accelerate fixes without waiting on engineers.
- McAulay argues autonomous agents are largely overhyped today; the real leverage comes from scheduled skills and human-steered, time-bounded agent work that closes loops and documents learnings.
- He emphasizes tool connectivity (PostHog, Stripe, CRM, email, Slack, analytics) and multi-agent critique to improve rigor, reduce mistakes, and speed up execution.
IDEAS WORTH REMEMBERING
5 ideasUse a goal cascade to stop AI from turning into a distraction machine.
McAulay’s system starts from an annual mission, cascades into quarterly OKRs, and then uses monthly/weekly retros and (optional) daily standups to keep work aligned; Claude is reminded of these goals at session start and is expected to push back on distractions.
Make retros data-driven by wiring Claude into your real systems of record.
Rather than relying on feelings (“sales seemed slow”), he connects tools like Stripe, PostHog, Xero, Attio, and Todoist so Claude can pull “hard numbers” during retros, spot patterns, and call out avoidance behaviors (e.g., overbuilding).
Treat growth work as reusable skills + scheduled checks, not one-off prompts.
The repo packages repeatable “skills” (SEO audit, PRD, YouTube packaging, tone of voice, retros) plus routines/cron jobs that can run periodically (e.g., weekly Lighthouse checks) to catch regressions and keep processes consistent.
“Autonomous” is useful when it means bounded, reviewable work—not magic money printers.
McAulay argues fully autonomous agents are mostly hype today; the practical win is human-in-the-loop agents that run bounded jobs (hours) or scheduled tasks (weekly/daily), then report back for review and decisions.
Replace expensive SEO suites with APIs + agents to get faster, cheaper audits and fixes.
He demonstrates an AI-driven SEO audit flow using Firecrawl (turn sites into markdown), DataForSEO (cheap keyword/competitive data), and Lighthouse CLI (speed/performance), producing agency-like audit outputs quickly and enabling rapid fixes via code agents.
WORDS WORTH SAVING
5 quotesAI has just completely unlocked what growth people can do. We no longer need to wait for engineers to build growth experiments. You can simply ship to production with permission from your engineers yourself.
— James McAulay
My current business hit that goal in ninety-five days. I haven't hired a single full-time employee.
— James McAulay
I think staying on track and I think, like, spotting your own patterns of maybe avoidance or procrastination or doing fun stuff instead of doing the stuff that's really gonna move the needle.
— James McAulay
The idea that we have these agents that we sort of create, and then they just go off and like print money for us without us sort of doing anything, I think is a little bit of a fantasy, um, and maybe some kind of overhyping on Twitter.
— James McAulay
I think people don't realize that you can ask an agent to improve itself.
— James McAulay
High quality AI-generated summary created from speaker-labeled transcript.
