The Straight-Answers Guide to Owning Your Own AI

1

The three ways to get AI in 2026

• Use a provider’s chat application. A chat app gives you access to the experience and models covered by that provider’s terms. Its account, subscription, and usage arrangements are separate from Fred.
• Own the harness, meter the intelligence. Buy or build assistant software that lives on your computer, and bring your own AI key so a provider bills you only for what you use. This is Fred's category — and honestly, also the category of free open-source setups if you're technical.
• Run a local model. Tools such as Ollama let Fred connect to models on your own hardware. You provide the model, hardware, and running setup.

Choose the software and model access that suit the work you want to do. Fred is a harness you connect to your own supported model access.

2

What a "harness" actually is — and why it's separate from the AI

The model (Claude, GPT, Gemini, a local Llama) is the thinking. The harness is everything that turns thinking into an assistant: which specialist handles your request, what tools it may touch, what it remembers about you, what it's allowed to do without asking, how it reaches you on your phone. When you buy assistant software, you are buying the harness — the intelligence inside it stays whoever's you choose. Any vendor who blurs that line is hiding either a markup or a lock-in.

3

BYOK, in plain words

Bring your own key means you set up access with an AI provider and connect that account to Fred using a supported key or access method. The provider bills you under its own terms. Fred’s software licence and the cost of model access and usage are separate.

4

The local-model truth

Local models run on hardware you provide, through tools such as Ollama. Their capabilities, memory requirements, and running costs depend on the model and your machine. Local inference is a model configuration choice within Fred.

5

Privacy has layers — ask where each thing lives

• Your files and documents — do they stay on your machine, or upload to the vendor's cloud?
• The assistant's memory of you — on your disk, or in their account system?
• Your credentials and keys — stored locally, or held by the vendor?
• Your conversations — here's the one honest vendors admit: with any cloud model, the words you exchange travel to that provider for thinking, on your key, under their data policy. Only a fully local model keeps even the conversation home.

A trustworthy product answers all four without flinching and tells you the fourth one straight.

6

Security questions that actually matter for an assistant that can act

A chat window can only say things. An assistant with hands — files, email, terminal — needs governance. Before buying anyone's (including ours), ask: Can I see and set per-agent permissions? Is there an approval gate before anything sensitive happens? Is there protection against prompt injection — content from the web tricking the assistant into acting against you — and does it block, or merely warn? If the answers are vague, the product is a demo, not a tool.

7

Plan for software and model access separately

Fred’s $99 price is the software licence. Your provider’s model access and usage are additional, ongoing costs under that provider’s terms. Estimate them from the models and workloads you intend to use, and track actual usage in the dashboard.

8

Red flags, from people who watch this market daily

• Check what model access includes. Understand the provider’s billing, usage limits, and terms alongside the price of any software you connect to it.
• Testimonials that can't be traced. Glowing quotes with stock faces. If they'll fake the reviews, what else?
• No stated limits. Every real product has a "what it doesn't do" list. Its absence is a statement.
• Category costume. Wrappers priced as platforms; chat apps sold as "agents"; agents sold as employees.
• Exit doors missing. No data export, no provider choice, no refund window.

9

The ten-question buyer's checklist

1. What exactly runs on my machine, and what runs on yours?
2. Where do my files, the memory, and my credentials live?
3. Whose AI does the thinking, and who bills me for it?
4. What happens to my assistant if I switch AI providers?
5. Can I see my real usage cost, live?
6. What can it do without asking me first — and can I change that?
7. What's the injection/security story, specifically?
8. What does it honestly NOT do?
9. What's the refund window, and the licence in plain words?
10. If the company disappears tomorrow, what still works?

10

Where Fred honestly fits

Fred is our answer to the middle path: a $99 one-time harness — eighteen configured specialist agents, a visual dashboard, an Android app, family rooms — that runs on your Windows or Linux machine with your own key, your files staying home. We think it's the best packaged version of ownership you can buy.

Fred is for people who want tools, workflows, and local context coordinated through the AI models they choose. The harness runs on Windows and Linux, with an Android app and WhatsApp or Telegram access from other phones. Your chosen model provider bills separately for access and usage.