startups-do

name: startups-do description: startups.do + Startups.Studio — define entire AI-generated startups as code, or operate them through the multi-surface control plane (CLI, API, SDK, MCP, web).

startups.do + Startups.Studio

You are an expert in startups.do and Startups.Studio — the platform for building startups as structured Business-as-Code objects, operated by a default team of AI agents across every surface.

When to Use

Activate when the user is defining a startup, operating a business via the platform, running a Foundation Sprint, generating startup ideas from a Thesis, or working with the Startups.Studio control plane.

Business-as-Code

A startup is a structured system, not scattered SaaS state. Define it in TypeScript:

import { Startup } from 'startups.do'
import { engineering, product, sales, marketing } from 'teams.do'
import { dev, sell } from 'workflows.do'

export default Startup({
  name: 'Acme AI',
  teams: {
    engineering,
    product,
    sales,
    marketing,
  },
  workflows: {
    build: dev,
    sell,
  },
  services: [
    'llm.do',
    'payments.do',
    'database.do',
    'org.ai',
  ],
})

That's a company. It builds products, sells them, and grows — operated by AI agents using the same agents.do SDK.

The Default Agent Team

Every new startup created via Startups.Studio ships with a core team of named AI agents pre-configured as agents records:

Agents and human roles are interchangeable in the type system — swap in a human with zero code changes:

import { priya } from 'agents.do'   // AI agent today
// import { sarah } from 'humans.do'  // hire Sarah tomorrow, same interface

The Multi-Surface Control Plane

Startups.Studio exposes one shared data model across five surfaces:

All surfaces read/write the same records. Move between them without rebuilding state.

Collection Families (50+ Collections)

Identity & Workspace: studios, startups, domains, users, templates

Discovery & Strategy: ideas, hypotheses, experiments, lean-canvases, story-brands, business-models, competitors, differentiators, approaches, theses, advantages

Model & Execution: nouns, verbs, things, functions, workflows, actions, tasks, routines, documents, events

Agents & Operations: agents, goals, projects, approvals, sandboxes, browsers

Commercial: customers, contacts, leads, deals, accounts, campaigns, products, services, subscriptions, budgets, cost-events, revenue-events

Data & Integrations: sources, resources, integrations, chats, messages

Agents Collection Schema

Every agent in a startup — AI or human — is a typed record:

{
  name: 'Priya',
  startup: { id: 'startup_abc123' },
  role: 'Product Manager',
  objective: 'Define what gets built and in what order',
  agentType: 'ai',           // 'ai' | 'human'
  adapterType: 'claude-code', // claude-code | agent-sdk | openclaw | cursor | gemini | human
  capabilities: ['product-planning', 'sprint-management', 'stakeholder-communication'],
  budgetMonthlyCents: 5000,
}

Foundation Sprint — Codified

Startups.Studio runs Jake Knapp's Foundation Sprint as CLI commands, producing a Founding Hypothesis:

If we help [customer] solve [problem] with [approach], they will choose it over [competitors] because our solution is [differentiation].

startups.studio discover customer    # → ICPs collection
startups.studio discover problem     # → Ideas collection
startups.studio discover advantages  # → Advantages collection
startups.studio discover competitors # → Competitors collection
startups.studio define hypothesis    # → Hypotheses collection

An AI agent can run the entire 2-day sprint in minutes. A human can work through it interactively.

Thesis-Driven Startup Factory

Define a strategic Thesis and the platform generates startup ideas at scale:

// 1. Define your thesis
startups.studio theses create --json '{
  "statement": "AI agents can replace professional services costing $50–200/hr",
  "domains": ["onet", "naics"],
  "filters": {
    "onet": { "minMedianWage": 50, "automationProbability": ">0.6" },
    "naics": { "sectors": ["54", "52"] }
  }
}'

// 2. Generate ideas (fan-out workflow)
// ~800 occupations × ~200 filtered industries = ~160,000 evaluations
// Each scored on TAM, feasibility, differentiation

// 3. Foundation Sprint on top ideas
// 4. Build, launch, measure

Reference data powering generation:

Platform Services

import { llm }      from 'llm.do'
import { payments } from 'payments.do'
import { db }       from 'database.do'
import { org }      from 'org.ai'
import { search }   from 'searches.do'

await llm`summarize this article`
await payments.charge(customer, amount)
await db.find({ collection: 'users', where: { active: true } })
await org.users.invite(email)

Bootstrapping a New Startup

# From idea to working startup
npx create-startups "AI-powered legal contract review for SMBs"

# What it does:
# 1. Registers org + startup in Startups.Studio
# 2. Provisions headless.ly entities (35 core + startup-specific)
# 3. Configures default agent team (Priya, Ralph, Tom, Rae, Mark, Sally, Quinn)
# 4. Runs Foundation Sprint (ICP, problem, advantages, competitors, hypothesis)
# 5. Returns CLI access + API keys + MCP endpoint

Best Practices

• Every startup is one shared model — CLI, API, SDK, MCP, and web all read/write the same records
• Start with the create-startups CLI for fastest path from idea to first authenticated request
• Agents and humans are interchangeable — design interfaces for both from day one
• org.ai manages identity and secrets — always provision it first
• Use the agents collection to define your team — both AI and human roles — before building workflows