What's the difference from using Claude Code, Cursor or Codex directly?

Direct is vibe coding — you send a prompt, get code, and quality drops as the project grows. N45.AI adds Spec-Driven Development: Discovery, Spec, Roadmap, parallel agents, automatic review and persistent state in the repository.

Do I need a Claude, Cursor or Codex subscription in addition to N45?

Yes. N45.AI orchestrates the workflow within the tool you already use. You need an active N45.AI subscription and at least one of Claude Code, Cursor or Codex.

Does it work with already-developed projects?

Yes. The Architect onboards existing code — detects stack, patterns and architecture — before any implementation. Brownfield, monorepos and microservices included.

Do I need to know how to write prompts?

No. You describe what you want in natural language. N45.AI talks with you, asks what's needed and orchestrates the execution.

What if the AI makes a mistake?

Automatic Code Review pipeline at every phase. The Architect validates diffs and acceptance criteria before advancing. You can intervene, pause or cancel at any time.

Can I pause and continue later?

Yes. State lives in files in the repository. Close the tool, come back days later — resumes exactly where it left off.

How does the trial work?

7 days free on the full Pro plan. Card required at signup. Cancel before the end and nothing is charged. Billing details on the pricing page.

Does N45.AI also handle infrastructure?

Yes. The DevOps Agent provisions infrastructure as code — Terraform, Pulumi, AWS CDK. Your production environment is born from the spec, not manual configuration.

Does it work with monorepos?

Yes. Tasks are generated by domain and executed in parallel by different agents, each in an independent worktree.

The workflow — Spec, Roadmap, tasks — lives versioned in your repository. Code generation uses the AI you chose, with the same privacy rules as that tool.

If I cancel, do I lose everything?

No. .n45/ files, specs and roadmaps stay in the repo. You lose access to agent orchestration, but the engineering history stays with you.

◆ Stop prompting. Start engineering.

Build complex applications. From zero to production.

N45.AI transforms Claude Code, Cursor and Codex into a real engineering squad: parallel agents, planned tasks and context that survives between sessions. From Discovery to Deploy, an interactive workflow you control.

Subscribe now →

◇ Claude Code◇ Cursor◇ Codex

Real output · Discovery → Spec → Code → Review → Deploy

◆ Who it's for

What's your scenario?

Indie Hacker

You develop solo — and need a full team.

Orchestrate agents with defined roles and move at squad speed.

Tech Lead

The team uses AI — but every dev ships differently.

One workflow, code generated from the Spec. The methodology standardizes delivery, not the prompt.

Startup Dev

Tight deadline. Code that needs to scale.

Discovery, spec and code in the same flow. Review before merge, zero surprises in production.

◆ How it works

Spec-Driven Development in an agents pipeline.

Discovery

Architect explores the domain, surfaces requirements, failure modes and trade-offs before any code.

Spec

Consultant agents create an approved technical document before any development. Zero ambiguity.

Code

Dev agents in parallel with independent worktrees implement the roadmap produced by the spec, following a controlled workflow.

Review

Architect agent automatically reviews each diff. Verifies acceptance criteria and completeness against the spec.

Deploy

DevOps agent provisions infrastructure as code and orchestrates CI/CD end to end — cloud or on-premises, your choice.

◆ Harness

The artifact that governs everything.

A living artifact, versioned in the repository, encapsulating execution rules.

Boundary

What is in and out of scope. Which modules can be touched, which data is accessible. The agent receives this as a constraint.

Contract

Expected inputs and outputs, mapped edge cases. The harness guides spec generation aligned with contracts without human intervention.

Guardrail

Business rules, performance limits, architectural patterns. Documented during onboarding of an existing application, and as features are generated.

Acceptance

Acceptance criteria defined before any code exists. TDD implementation by default.

Repository-generated documentation

When creating a project, starting a feature, or working in an existing codebase, N45.ai generates and maintains versioned documentation in the repository — the living context that guides every agent.

Documentation

Project

Project identity: type, description, and high-level scope. Created during onboarding and evolved with the product.

Patterns

Architectural conventions, code patterns, observability, and testing. Guardrails every agent respects.

Stack

Technical map: languages, frameworks, dependencies, and layers (frontend, backend, infra).

Runbook

How to install, run, test, and validate health. Commands, ports, and local URLs.

Changelog

Delivery history by feature and fix, updated on every merge.

Tech Debt

Technical debt identified during discovery and execution, with status and traceability.

Diagrams

Architecture Diagram

Structural view: components, layers, and how they connect.

Flow Diagram

Process flows and interactions between parts of the system.

Context Diagram

C4 context view: actors, external systems, and project boundaries.

◆ Agents10 agents

A squad of agents, not a prompt.

Generic AI becomes an engineering squad: each agent has a defined role, isolated context and traceable responsibilities.

Architect

Discovery, spec and automated review

Frontend

UI, components and visual integration

Backend

APIs, services and business logic

DBA

Schema, queries and migrations

DevOps

Infrastructure as code and CI/CD

QA

Tests and acceptance criteria

UX/UI

Flows, wireframes and accessibility

Security

Threat modeling and guardrails

SRE

Observability, SLAs and resilience

Data Architect

Data modeling and pipelines

◆ SDD vs Vibe coding

Vibe coding doesn't scale.

You describe an idea, AI generates code, you test, tweak, change the prompt, paste pieces together, initial patterns are forgotten, quality drops drastically — you persist until it works.

DimensionVibe CodingN45.AI · SDD

Starting pointVague idea → code✓Discovery → Spec → code

ArchitectureDecided at the start → forgotten after new prompts✓Documented in Spec → reviewed and enforced at every agent run

Cross-session consistencyLow — every compaction, important definitions are forgotten✓State lives in files

Output qualityVariable, depends on the prompt and context size when starting work✓Automatic code review per phase

TraceabilityNone✓Spec + Roadmap + Tasks + Changelog

For long projectsBecomes chaos quickly✓Scales — controlled workflow in Phases

◆ SUCCESS RATE × SCALE

N45.AI × Vibe Coding

Vibe coding nails the MVP and collapses as the project grows — the entropy breaking point. N45.AI keeps the success rate steady at any scale.

N45.AI

Vibe Coding

◆ Next step

Scale without losing quality.

Vibe coding drops off as projects grow. N45.AI keeps your workflow stable at any scale.

Subscribe now →View pricing

◇ 7 days free on Pro · new subscribers only◇ Install via npx · get started in a few minutes

How to use

Get started in 5 minutes:

Compatible with Windows, Linux and Mac. Get running quickly via npx.

Prerequisites

Node 22+GitN45.AIClaude Code / Cursor / Codex

Step by step

1

Authenticate machine

npx n45-ai login

2

In the project directory

npx n45-ai install

Select your tool

3

Run in the AI prompt

Ready

/n45

/n45Entry point — detects context and routes to create, fix, continue, or status

/n45-featPlans and implements a feature or project, from discovery to execution

/n45-fixInvestigates and fixes bugs, errors, or unexpected behavior

/n45-statusShows roadmap progress: active phase, completed/pending tasks, and blockers

/n45-nextResumes a paused roadmap and continues pending tasks

/n45-runStarts or stops the application (dev server, containers, local services)

Go to documentation →

◆ Integrations

Runs with your tools.

N45.AI orchestrates on top of what you already use. No lock-in, no stack migration.

Claude Code

by Anthropic

● Available

Cursor

by Anysphere

● Available

Codex

by OpenAI

● Available

Use cases

The same pipeline. Four real contexts.

Discovery, Spec, Code, Review and Deploy — from zero to production, with validation at every phase.

New app

SaaS booking app

From zero to deploy — patterns, stack and a controlled roadmap through to production.

Discovery

Domain, personas and booking flows mapped before any code.

Spec

Stack, patterns and acceptance criteria documented and approved.

Code

Phased roadmap — auth, calendar, payments — agents in parallel.

Review

Diffs reviewed against spec; tests and acceptance validated automatically.

Deploy

CI/CD provisioned; staging → production deploy.

Between each phase, you validate progress before moving forward.

◆ Contact

Ready to get started?

Talk to us and schedule a demo of how N45.AI can accelerate your team.

Information

✉

CORPORATE EMAILcontato@n45.ai

◎

PHONE / WHATSAPP+55 11 97105-9034

◈

LOCATION

São Paulo, SP — Brazil

◆ Real questions

Get your questions answered before subscribing.

Build real systems,
not throwaway prototypes.

Subscribe now →

Build complex applications. From zero to production.

What's your scenario?

You develop solo — and need a full team.

The team uses AI — but every dev ships differently.

Tight deadline. Code that needs to scale.

Spec-Driven Development in an agents pipeline.

Discovery

Spec

Code

Review

Deploy

The artifact that governs everything.

Boundary

Contract

Guardrail

Acceptance

Repository-generated documentation

Project

Patterns

Stack

Runbook

Changelog

Tech Debt

Architecture Diagram

Flow Diagram

Context Diagram

A squad of agents, not a prompt.

Architect

Frontend

Backend

DBA

DevOps

QA

UX/UI

Security

SRE

Data Architect

Vibe coding doesn't scale.

N45.AI × Vibe Coding

Scale without losing quality.

Get started in 5 minutes:

Devs who stopped prompting.

Runs with your tools.

The same pipeline. Four real contexts.

SaaS booking app

Discovery

Spec

Code

Review

Deploy

Ready to get started?

Get your questions answered before subscribing.

Build real systems,not throwaway prototypes.

Build real systems,
not throwaway prototypes.