Your code never leaves your Mac: private AI for client work
Private AI development means running a local model where nothing leaves the machine, the honest way to use AI on client code under NDA. Here's how it works.
Private AI for client work means running the model on your own machine, so the client’s code, content, and credentials never leave it. If you build for clients under NDA, that’s the difference between using AI and being contractually unable to. ReDock is a free native macOS app that makes it practical: a built-in agent, a local-LLM catalog sized to your Mac’s RAM, and a full local site platform — with a local model, nothing goes to the cloud.
The problem with cloud AI on client code
Cloud models are capable, but using one means your inputs leave your machine to be processed on someone else’s servers. For a lot of work that’s fine. For client work it often isn’t.
Client contracts and NDAs frequently restrict where code and data can go. “We send the client’s proprietary codebase to a third-party API” is a sentence that ends engagements. Even when it’s technically permitted, it’s a disclosure and a risk you have to reason about. And it’s not just source: it’s database contents, customer data in a staging site, API keys in a config file. Anything the agent can see is something that would be transmitted.
The safe move for sensitive work is to not transmit it at all.
What “nothing leaves the machine” actually means
With a local model, inference happens on your Mac. The model weights live on disk, the computation runs on your hardware, and the prompt and response never touch a network. There’s no API endpoint in the loop because there’s no API.
ReDock builds the whole workflow around that:
- Local models, on-device. The built-in catalog is tiered to your Mac’s RAM and runs models locally — no API keys, no accounts, no per-token metering. See the built-in agent docs.
- Local sites. WordPress, headless WP, Astro, Next.js, Node, Bun, PHP, and MySQL all run on your machine on real
https://name.testdomains. The agent operates them over alocalhostMCP endpoint. - You control exposure. Nothing is shared until you deliberately open a Cloudflare tunnel or deploy. Client code stays put by default.
Put together, an agent can build, edit, and verify a client’s site with a model that never sends a byte off the device.
Where the tradeoff is honest
Local models don’t match large cloud models on the hardest reasoning — Claude vs local models for agent work: an honest comparison lays that out plainly, and Local LLM on 8GB Macs covers what fits on a modest machine. For client work the calculus is different from a personal project: for anything covered by an NDA, “runs locally” can outweigh “reasons slightly better,” because the cloud option isn’t on the table at all.
And a lot of client work is exactly what a local model does well: bounded, tool-driven tasks where ReDock’s MCP tools carry the correctness and the model just picks the call.
What driving it looks like
You’re under NDA for a client’s WordPress build. You pick a local model from ReDock’s catalog and point the built-in agent at the client’s local site. You ask it to add pages, wire a contact form, and restyle the header. It reads the site’s patterns, makes the calls, and ReDock executes and verifies each one. The client’s content and code stay on your Mac the whole time. When the work’s approved, you deploy on your terms. For the WordPress specifics see WordPress MCP: let Claude build, edit, and manage WordPress for you and the pillar No API key AI: run a local model that can actually do things.
FAQ
Does a local model guarantee nothing is sent anywhere?
When you use a local model, inference runs on your Mac and the prompt and response don’t touch a network. If you later choose to connect a cloud model or share a site, that’s a deliberate step you take — the local default keeps things on-device.
Is a local model good enough for real client work?
For bounded, tool-driven tasks, yes — and ReDock’s MCP tools verify each change. For the hardest reasoning you’d want a cloud model, but for NDA-covered work the local option is often the only one you can use.
Do I need to configure keys or accounts?
No. Local models run without API keys or accounts. You pick one from the catalog and it runs on your machine.
If you want AI on client code that stays on your Mac, download ReDock — one download and a five-minute install.