Lazynext — The Autonomous Company Operating System

Lazynext is the Autonomous Company Operating System for individuals, professionals, teams, startups, and organizations. You create a Company and set its direction — mission, strategy, goals, KPIs. Lazynext plans, executes, measures, learns, and continues the work autonomously, within your budget, permission, and approval policies. The platform provides identity, organizations, workspaces, projects, tasks, documents, files, automations, a 12-role AI agent workforce with a durable autonomy loop, an approval center, multi-level budgets, company memory, CRM, customer support, finance, research, product planning, opportunity detection, integrations, analytics, search, a developer API, an MCP server, and a Creative Studio module for AI ad generation powered by Atlas Cloud.

One environment for your entire company. You set the direction — Lazynext runs the work. Identity is unified, data is unified, permissions are unified, and every module shares the same platform core.

What Lazynext is

Lazynext is the Autonomous Company Operating System. The core entity is the Company: mission, vision, strategy, goals, KPIs, plans, products, customers, and memory. Around it runs a durable autonomy loop (observe → understand → prioritize → plan → select agent → select tools → check policies → approve → execute → verify → measure → record → update memory → replan) with four autonomy modes — Manual, Assisted, Autonomous, and Timed Continuous — and a 12-role AI agent workforce (CEO, Strategy, Research, Product, Engineering, Design, Growth, Sales, Support, Finance, Operations, Security) whose actions pass an 8-layer permission evaluator, budget controls, and an approval center.

The shell wraps everything in one coherent platform UI: each module is a first-class application with its own navigation, data, and actions, but all modules share the OS shell, identity layer, permission system, workspace context, search, audit, memory, command palette, and design system.

Platform modules

Platform primitives

All modules build on shared domain concepts — 143 Prisma models in total: User, Identity, Session, Organization, Company, Goal, Kpi, Plan, Workspace, Team, Membership, Role, Permission, Project, Task, Document, File, Asset, Conversation, Integration, Automation, Agent, ToolDef, ToolCall, Approval, Budget, BudgetEntry, Memory, Event, DetectedOpportunity, Recommendation, ApiCredential, WebhookEndpoint, Notification, AuditEvent, Subscription, Invoice, UsageRecord, and ScheduledJob, among the full business-domain model set.

Creative Studio

The Creative Studio module (the Growth domain of the Company OS) preserves the original ad-creative product value within the OS framework. It includes:

• UGC Product Ad — Product + presenter photos → lip-synced UGC ad

• Reference to Ad — Upload a viral ad → remake it with your product

• AI Drama Ad — One topic → comedy script → shot-by-shot drama ad

• Ad Skit — One-line product → two-person comedy skit

• 150+ creative tools — Hook generators, brand voice analyzers, A/B testing, performance forecasting, competitor intelligence, compliance checking, and more

All workflows auto-detect input language and support 13 locales.

Related MCP server: Omni-Video Studio MCP

Who it is for

• Founders and solopreneurs who want an AI workforce to operate substantial parts of their company

• Startups and teams who want a unified autonomous workspace instead of fragmented tools

• Organizations that need shared identity, permissions, budgets, approvals, and data across modules

• E-commerce teams and agencies that need AI ad creative production within a broader platform

• Developers who want a public API and MCP server for programmatic access

• Businesses that need audit trails, compliance, and enterprise-grade security

Quick start

git clone https://github.com/Lazynext-Platform/lazynext.git
cd lazynext
npm install

For local development:

cp .env.example .env.local
# Fill auth, Atlas Cloud, Dodo Payments, and Resend values.
npm run dev    # starts on port 3100

For Cloudflare/OpenNext preview:

cp .dev.vars.example .dev.vars
npm run cf:preview

Core application variables:

ATLASCLOUD_API_KEY=
NEXTAUTH_SECRET=
NEXTAUTH_URL=http://localhost:3100
AUTH_URL=http://localhost:3100
GOOGLE_CLIENT_ID=
GOOGLE_CLIENT_SECRET=
ADMIN_EMAILS=support@lazynext.com

Platform resources:

Where to get each value:

Open http://localhost:3100.

Mock Atlas Cloud API (for local testing)

npm run mock-atlas    # starts on port 3099

The mock server returns realistic LLM responses, simulates generation task lifecycle, and serves placeholder media — without consuming real credits or making external API calls.

Deploy

Lazynext deploys exclusively to Cloudflare Workers using OpenNext.

npm run cf:deploy

Cloudflare uses the DB D1 binding and MEDIA_BUCKET R2 binding in wrangler.jsonc. The deployment includes a custom domain route for lazynext.com.

Set application secrets with Wrangler or the dashboard:

wrangler secret put ATLASCLOUD_API_KEY
wrangler secret put NEXTAUTH_SECRET
wrangler secret put GOOGLE_CLIENT_ID
wrangler secret put GOOGLE_CLIENT_SECRET
wrangler secret put TOKEN_ENCRYPTION_KEY
wrangler secret put CRON_SECRET

See DEPLOYMENT.md for resource creation, database initialization, and verification commands.

Developer Platform

Public REST API (v1)

Lazynext exposes a versioned REST API at /api/v1/ with API key authentication, scopes, and rate limits. Endpoints cover workspaces, projects, tasks, and documents.

• API key management at /settings → Developer

• OpenAPI-style documentation at /developers

• Rate limits: 60 req/min (standard), 10 req/min (AI endpoints)

MCP Server

Lazynext includes a production-ready MCP (Model Context Protocol) server at /api/mcp implementing the 2026-07-28 protocol specification:

• Stateless core (no initialize/handshake)

• Streamable HTTP transport

• OAuth 2.1 protected-resource metadata

• Tools: list_workspaces, get_workspace, list_projects, create_task, list_tasks, search, and more

• Every tool requires authentication and workspace scoping

• MCP metadata at /.well-known/oauth-protected-resource

Credits and pricing

Users spend in-app credits while the deployment pays Atlas Cloud in USD. Credit packs are available at $9/$39/$99 with ~30 display currencies.

src/lib/video-pricing.ts calculates video credits:

credits = ceil(perSecond[resolution] * seconds * ACCOUNT_MARKUP * MARGIN / CREDIT_USD)

Top-up packs live in src/config/pricing.ts.

Testing

npm test           # 7,750+ unit tests
npm run test:e2e   # 1000+ E2E tests (Playwright)
npm run lint       # ESLint (0 errors)
npm run build      # Production build

Technical architecture

lazynext/
├── src/
│   ├── app/                    # Next.js 16 App Router (435+ page directories, 3,290+ API routes)
│   │   ├── (Company OS)/       # company, goals, plans, agents, autonomy, approvals, crm, finance…
│   │   ├── creative/           # Creative Studio module (Growth domain)
│   │   ├── api/v1/             # Public REST API
│   │   ├── api/mcp/            # MCP server (2026-07-28)
│   │   └── api/                # Internal API routes
│   ├── components/             # Reusable UI components
│   ├── config/                 # Navigation, pricing, app catalog
│   ├── i18n/                   # 13 locales with dynamic loading
│   └── lib/                    # 250+ domain service modules, agent runtime, security, providers
├── prisma/schema.prisma        # 143 models, D1/SQLite
├── e2e/                        # Playwright E2E tests
├── test/                       # Unit tests
├── docs/adr/                   # 222 Architecture Decision Records
├── docs/transformation/        # Autonomous Company OS transformation docs & acceptance
├── research/                   Discovery + architecture reports
├── auth.ts                     NextAuth v5 config
├── open-next.config.ts         OpenNext build target
└── wrangler.jsonc              Cloudflare deployment config

Stack

License

MIT. See LICENSE. AI generation powered by Atlas Cloud.

Lazynext is a Lazynext-branded distribution built on the open-source Atlas Marketing Studio project (MIT). All AI generation runs through the Atlas Cloud API.