Skip to content
Book a review

NEW YORK · PRODUCTION READINESS FOR AI-BUILT SOFTWARE

We finish what AI started.

You built it with Lovable, Bolt, Replit, Cursor or Claude Code, and people are using it. We make it secure, reliable and ready to grow, without asking you to stop building with AI.

Book a review ($4,500)Run the free check
AIAGENTSMOBILEWEBDESKTOPAPI

Select a system to explore

WHAT WE SEE

AI gets you to launch. It doesn't get you to production.

The same problems show up in almost every AI-built app.

  • Open data

    Database rules that let any signed-in user read everyone's records.

  • Exposed keys

    API keys shipped in the browser bundle where anyone can copy them.

  • No safety net

    No tests, no backups, no monitoring. You find out about outages from customers.

  • Runaway costs

    AI and hosting bills that grow faster than revenue.

  • Code nobody owns

    A codebase that works, but nobody can explain or safely change.

How we work

Start small. Every step stands on its own.

  • FREE

    Live-site check

    Free

    A quick check of your live site’s public surface. Verify your domain, get results in minutes.

    • Exposed keys in the bundle
    • Open database rules
    • Missing security headers
    • Auth gaps
    Run the free check: Live-site check

  • START HERE

    Production Readiness Review

    $4,500, 5 days

    A senior review of your code and infrastructure: security, data, auth, cost, performance and maintainability. You get a scored report and a ranked fix list with estimates.

    • Scored report
    • Ranked fix list with estimates
    • Not a penetration test
    Book a review: Production Readiness Review

  • MOST COMMON

    Harden Sprint

    $15k–$40k, 2–4 weeks

    We fix what matters most and add tests, CI, monitoring, backups, rate limits and cost controls. Fixed price.

    • Critical fixes
    • Tests, CI and monitoring
    • Backups, rate limits and cost controls
    See the Harden Sprint: Harden Sprint

  • WHEN YOU GROW

    Scale Build

    Proposal

    A native mobile app, a public API, a desktop client, agents, or rebuilding the parts AI got wrong.

    • Mobile, web, desktop, API or agents
    • Rebuilds where AI got it wrong
    • A codebase we already know
    Request a proposal: Scale Build

  • ONGOING

    Production Care

    From $2,500/month

    On-call support, patches, cost monitoring, and a senior engineer reviewing every AI-written pull request.

    • On-call support and patches
    • Cost monitoring
    • Review of every AI-written pull request
    See Production Care: Production Care
  1. 01CheckMinutes
  2. 02Review5 days
  3. 03Harden2–4 weeks
  4. 04ScaleProposal
  5. 05CareMonthly

WHEN YOU'RE READY TO GROW

One team, every surface.

We already know your codebase, so the next surface is faster.

  • Mobile

    iOS and Android apps built on the product you already have.

  • Web

    Hardened web apps, portals and dashboards that hold up under real traffic.

  • AI

    Desktop

    Mac and Windows software that works with your files, documents and tools.

  • AI

    API

    Public and partner APIs, versioned, documented and rate-limited.

  • Agents

    Agents that handle multistep work across your systems, with people in the loop.

Why Castro AI

  • We like AI-built code

    You were right to build it with AI. We harden what you have instead of opening with "rewrite everything".

  • Fixed prices, published

    The review is $4,500. Sprints are fixed-price. You know the cost before we start.

  • AI-native, so fast and affordable

    Our own tooling does the repetitive checks, so senior time goes where it matters.

  • Senior engineers only

    The people you meet are the people who do the work.

  • You own everything

    Code, data, designs and infrastructure accounts are yours.

  • Based in New York, working across the US

    In person in New York, remotely across the US.

Selected work

All case studies

Questions

  • No. Keep building with them. Production Care adds a senior review on every change so speed doesn’t cost you stability.

  • No. It’s a senior engineering review of security, reliability, cost and code health. If you need a certified penetration test, we’ll tell you and refer you.

  • Apps built with Lovable, Bolt, Replit, Cursor, Claude Code and similar tools, most often on React, Next.js, Supabase, Firebase and Postgres. [PLACEHOLDER: confirm list.]

  • Only the parts that need it, and only if you agree. Most apps need hardening, not a rewrite.

  • You do: code, data, designs and infrastructure accounts.

  • Yes. In person in New York, remotely across the US.

People are using what you built. Let's make sure it holds.

A Production Readiness Review takes five days and costs $4,500.

Book a review

NEWSLETTER · WEEKLY

The Build Brief

A weekly teardown of one AI-built app: what broke and how we fixed it.