10 specialized Claude Code sub-agents modeled on the actual Septim Labs executive team we use to run a solo-founder SaaS studio. Atlas (chief of staff), Luca (CTO), Canon (brand), Ember (CMO), Tally (CFO), Nova (design), Ward (legal), Mira (customer), Juno (research), Pip (intern). Drop into ~/.claude/agents. Invoke with /agents. Ship faster.
Each agent has a distinct voice, scope, tool set, and judgment-bias. Use Atlas to plan, Luca to build, Canon to name, Ember to market, Tally to price, Ward to derisk, Nova to design, Mira to humanize, Juno to research, and Pip for the grunt work nobody wants.
Ruthless prioritization. Writes sprint plans, triages, cuts scope. Proactively used before any multi-week build.
Architecture, feasibility, stack selection, performance audits. Tells you honestly when a promise exceeds what a solo operator can deliver.
Names products, enforces brand architecture, approves copy. Has taste and uses it. Won’t greenlight derivative.
Positioning, launch sequencing, landing copy, ICP refinement, content calendars. Thinks in funnels, not press releases.
Pricing tiers, unit economics, cost modeling, runway calculations. Says “that free tier will cost you $14/user/mo” before the bill arrives.
UI/UX reviews, visual hierarchy, animation polish, interaction design. Calls things ugly when they’re ugly.
ToS, privacy policies, contractor agreements, compliance flags. Knows when a real attorney is needed and says so.
Onboarding sequences, support escalations, customer-facing microcopy. Voice of the user inside every meeting.
Market research, competitor analysis, Reddit/HN signal, industry reports. Returns structured findings with sources.
The 5-minute tasks that distract specialists. Summarizing, format conversion, quick fact-checks, tidying messy lists.
The pack is a directory of markdown files that live under ~/.claude/agents/septim/. Each file defines an agent’s persona, model tier, tool allowlist, and proactive-use triggers. Claude Code reads them automatically and surfaces them via the /agents command.
That’s it. No config. No env vars. No API tokens. The agents run on your existing Claude Code subscription — zero extra cost.
Secure Stripe checkout · land on a short form (email + GitHub username) · repo invite within 24 hours.
Yes. The agents run on Claude Code (free) or Claude Max (paid). They use your existing subscription for inference — we don’t resell tokens. If you’re already using Claude Code, zero additional cost.
Yes. This is the same roster we run Septim Labs with every day. Two of our products (Forge and Gauge) were scoped, designed, and shipped largely by this exact team working in parallel.
Sub-agents have isolated context, dedicated tool allowlists, persistent personas, and run in parallel. You can fire Atlas, Juno, and Luca at the same problem simultaneously — each returns their domain view without polluting the others’ context. That’s the unlock.
Absolutely. They’re plain markdown. Fork, tune the voice, add tools, switch models. That’s the point — they’re a starting point for your exec team, not a sealed product.
Then this is the natural next purchase. The Kit gives you the infrastructure. The Agents give you the team. Both together = a one-person studio that ships.
30 days. Email us, we refund. No friction. If the agents aren’t actively saving you time, we’d rather give the money back.
10 agents, each with their own voice and job. $49 once. Your Claude subscription powers the rest.
Buy the Agents Pack — $49 lifetime →