Easily manage components and releases.
Keep firmware, APIs, iPhone, and Watch builds together. Review a version, test each component, then publish the release set.
From web apps to connected products, Millwright builds, ships, and operates production systems on real infrastructure.
Run on your Mac, Drydock Cloud, or your own servers. Build firmware and an API alongside signed iPhone + Apple Watch apps with TestFlight delivery.
free•Mac app + web + CLI
Mac app: Apple silicon + macOS 14 or later
Google, GitHub, email, or passkey
Describe the product. Pick the pieces, or let the tiles light up as you type.
Start new, bring a repo, or connect a live host. Millwright handles the build, release, and infrastructure around it.
Keep firmware, APIs, iPhone, and Watch builds together. Review a version, test each component, then publish the release set.
Describe it, import a GitHub repo, or connect an existing server. Keep Cursor, VS Code, Claude Code, or Codex; Millwright discovers the components and takes it from there.
Local Preview runs the real app on your Mac. Browser, behavior, security, and performance checks land as readable reports. Siege checks security and dependencies, Regatta compares performance, and Shakedown checks behavior against the spec.
Homeport publishes what runs on your Mac. One click moves it to an always-on Drydock Cloud VM. Easily add existing servers to your Millwright fleet.
Review the runtime, framework, database, hosting, and topology before the agents build. See the alternatives and shape the engineering plan alongside the design.
Bring ESP32 firmware, iPhone + Watch apps, backend, and infrastructure together. Flash boards, enroll devices, and promote firmware updates over the air.
Provision a new VM or connect an existing host. Millwright keeps releases, checks, TLS, DNS, usage, and live topology together.
Full root and SSH, public IPv4 + IPv6, and 45+ locations. Ordinary Linux processes run on loopback ports behind Caddy, with preview hostnames and a stable live URL.
Every build gets its own preview URL. Pin component versions into a release set, review their checks, and promote a known product state. Retained, checked deployments give you a path back.
Every service, port, route, and datastore in the project, drawn as it actually runs — from Local Preview on your Mac to the multi-region deployment.
Hosts, Homeport routes, usage meters, and check reports, side by side. Add cloud VMs, on-prem GPU servers, or spare Mac minis to one fleet.
ddsh is an AI shell for every node: say what you need and it runs the right commands on the right machine. Plain bash still works.
Organizations bring the whole team into one project — shared context, owner approvals on pending checkpoints, and scoped GitHub integrations and scoped MCP instead of shared credentials. What the agent learns in one project carries to the next through the shared memory layer.
Your code stays in your GitHub repo, and Drydock Cloud VMs stay ordinary Linux machines you can inspect and operate directly — no proprietary runtime between you and the product.
The local workflow is free, and Homeport publishing includes 100K requests / mo and 100 GB / mo of relay bandwidth. Pay when you add cloud.
Connect Claude Pro or Max or ChatGPT Plus or Pro, or use an Anthropic or OpenAI API key — billed direct, nothing on top. A local model option (OpenCode + Qwen) is coming soon.
An always-on VM when the project outgrows your Mac — 1 vCPU · 1 GB · 25 GB. Larger VM sizes are available; pricing and inventory vary by region.
Hosted model access with metering, balances, and spend controls. Until then: your subscription, your key, or a local model.
Managed edge compute, storage, cron, and managed backends — pricing arrives with the release.
Private branch drafts, parallel release lines, merge queues, analytics, structured logs, and audit records.
Expanded MCP integration, cross-project activity, and scheduled agents.
Managed compute and data services, multi-region hosting, and Android with Google Play delivery.
Expanded local models and offline workflows, plus hosted inference with usage visibility and spend controls.
The local workflow — build, run, preview, publish — is free, and Homeport includes 100K requests and 100 GB of relay bandwidth per month. Cloud VMs start at $5/mo when you want one.
No. If you have Claude Pro/Max or ChatGPT Plus/Pro, connect it and Millwright uses Claude Code or Codex — or bring an Anthropic or OpenAI API key. A local model option (OpenCode + Qwen) is coming soon.
Millwright for Mac requires Apple silicon and macOS 14 or later. Local model support is coming soon and will want at least 32 GB of RAM. On an Intel Mac, Windows, or Linux, use Millwright Web to build on Drydock Cloud with Claude Code, Codex, or an API key. Hosted inference credits are planned.
Yes — Homeport gives a public URL to the app running on your Mac, so it serves while the Mac is online. Move the project to an always-on VM when it shouldn’t depend on your laptop lid.
Yes — keep vibe coding in Cursor, VS Code, or your own Claude Code. Millwright runs the infra, and every push rebuilds.
Yes — import a GitHub repo or connect an existing server without rebuilding the system from scratch.
In a GitHub repo you own. Cloud VMs — and dedicated bare metal by request — come with full root and SSH, with no lock-in layer between you and your product.
A web app, API, standalone CLI, and an iPhone app with a Watch companion — native SwiftUI with a live Simulator preview, installed on your own iPhone over USB, then delivered through TestFlight and the App Store with App Review pre-flighted and screenshots and privacy pages prepared. Connected devices and robotics belong in the same project, with ESP32 flashing, enrollment, and OTA delivery.
No — the same project opens from the web app and the standalone CLI. The Mac app is the closest seat to a local build.
Download the Mac app, or start in the browser — it’s the same project either way.
Mac app requires Apple silicon and macOS 14 or later.
Want an invite code for a free cloud VM? Ask us on Discord →