Official GitHub presence for HyperLabs, a venture studio that co-founds and scales digital companies — Launch (GTM & growth), Build (product & AI), Transform (digital reengineering). We work with skin in the game: operating, technical and strategic capital, not only money.
Here you'll find two main types of repositories:
- Open source tools & SDKs, MIT-licensed and published to npm. These are the pieces we extracted from our own products because they were useful outside them.
- Product repositories, which back the platforms we build and operate. Most are private; access is granted to the team and to partners under an engagement.
If a repository's status is unclear, treat it as private product code and reach out before use.
Leer en español: README.es.md
Everything here is MIT and runs in production inside one of our products first.
| Repository | What it is |
|---|---|
| russell | A declarative language for conversational agents. Node topology, intent routing, per-node model & temperature, deterministic guards, escalation policy and persona — as versioned, portable JSON a runtime compiles and executes. No code, no redeploys. → russell-schema |
| hyperflow-sdk | HyperFlow's agent layer for any app. Google Ads audits, SEO, strategy and conversational commerce with live progress, sessions and guards in a few lines of TypeScript. → hyperflow-sdk |
| roz | An intelligence layer over engineering work: watches what happens on GitHub, documents it, routes it and notifies with context — autonomously. → roz-ops.vercel.app |
Issues and pull requests are welcome on these three. Read the repository's own README first — each one states the scope it accepts.
Not a library: a starting point. Install the skills you need and leave the rest.
| Repository | What it is |
|---|---|
| hyper-skills | Our skills for AI coding assistants (Claude Code, Cursor, …): multi-tenant backend architecture, SDK design, RAG and agent patterns, documentation, analytics integration. One folder per skill. |
npx skills add hyperlabs-ai/hyper-skillsThe platforms the studio builds and operates. Their repositories are private.
| Product | What it does |
|---|---|
| Orwel | End-to-end analytics and attribution: follows a visitor from the ad to the revenue, computes their profile and exposes everything over MCP. |
| HyperFlow | CRM with AI agents: prioritizes leads, drafts follow-ups, schedules, quotes and issues the CFDI. |
| TeamUp | AI candidate evaluation: stratified scoring, a single rubric and an interview guide. |
| Watersy | Online store for reverse-osmosis and water-treatment equipment: catalog, cart and checkout, on React and Vite. |
Every product ships with its own design system, taken from its code rather than from a preset — the tokens, the type scale and the composition rules live next to the product they belong to. The stack underneath is consistent: TypeScript, SvelteKit and Next.js, Postgres via Supabase, Vercel, and an AI layer built on retrieval and agents rather than on a single prompt.
Two rules survive every project: a readable artifact beats a diagram — a real commit, a real invoice, a passage with its provenance — and numbers are derived, never hand-written, so they don't lie when the data around them changes.
- Site — hyperlabs.vc
- Applications and partnerships — the form on the site
- Security or anything that shouldn't be public — open a private security advisory on the repository in question
Repositories not listed above are client engagements or work in progress. Ask before you build on them.
