ProbahiCraft Agents
Your AI agent, on your site. The infrastructure is ours.
You describe what you need in a chat. We assemble the agent, host it, and give you a snippet to paste into your site. It answers in your brand, on your pages. Everything behind it — the reasoning, the memory, the tool calls, the updates — runs on ours.
There's no signup yet. The waitlist is an email to a person, and we'd rather it stayed that way until the product is ready.
What it actually is
Most businesses that want an AI agent have the same two problems. They can't describe what they want in the language a developer needs, and even if they could, they don't want to run AI infrastructure to keep it alive.
ProbahiCraft Agents removes both. You talk to an intake agent in plain language until the scope is agreed. We build the agent from that. It lives on your site, under your brand, and we run the machinery underneath it — models, memory, tool connections, updates — so none of it becomes your operations problem.
The short version: you get to offer your customers a custom AI agent without becoming an AI company to do it.
Four steps, and none of them are yours to build.
Tell it what you need
You chat with an intake agent. It asks the questions a good technical brief would ask, and it doesn't stop until the scope is clear. Nothing gets built until you've confirmed what it wrote down.
We assemble the agent
The builder turns that brief into a working agent: the role it plays, the tone it takes, what it's allowed to do, what it remembers, and which of your systems it's connected to.
You test it, then paste it in
You get a live sandbox link before anything touches your site. When you're happy, you paste one snippet — a script tag or an iframe — into your pages. It's a copy-paste job, not an integration project.
It runs, and we keep it running
Every conversation is handled on our infrastructure. Model upgrades, memory, tool calls and fixes happen on our side. You're billed for the capacity you use.
What it can be wired into
An agent is only useful if it can see the things it's talking about — the order, the stock, the customer record. These are the connections we're building first.
If what you run isn't on that list, it probably speaks REST or webhooks, and that's the fallback.
You don't have to build any of this.
This is the part that usually stops a business from having its own agent. It's also the part you never have to see.
Hosting and isolation
Cloud infrastructure that scales with the traffic, with each customer's data kept apart from every other customer's, plus the accounts and billing around it.
Models and retrieval
A capable model for the hard turns, a lighter one for the easy turns, and a retrieval layer so answers come from your material instead of the model's imagination.
Orchestration
The part that decides what a request actually needs: which step comes next, what to remember, which tool or API to call, and when more than one agent should be involved.
Delivery and versioning
Every agent is versioned and deployable, so a change ships or rolls back without taking your site down. Usage is metered as it happens.
Your layer
The embed on your site, an API and SDK if your developers want to go further, a sandbox to try things in, and the option to run it entirely under your own branding.
How it'll be sold — as far as we've decided.
We're not publishing prices for a product nobody can buy yet. What we can tell you is the shape.
- Starter
- One agent, a modest monthly conversation allowance, one integration, sandbox and embed included. Enough to find out whether this works for you.
- Growth
- Several agents, a much larger conversation allowance, every integration we support, and priority support. Meant for a business running this properly, not trialling it.
- Scale
- A larger fleet of agents, a high conversation allowance, custom integrations and webhooks, and the option to run it fully white-labelled.
- Enterprise & white-label
- Dedicated infrastructure, a service-level agreement, no cap on agents, and a reseller arrangement for agencies who want to offer this to their own clients. Priced per case.
No prices here yet, on purpose. A number we'd have to revise is worse than no number.
Where this is right now
In development. There's no public signup, no free trial, and no launch date we're willing to put in writing.
What we do have is a clear design, a team that has built AI systems for clients before, and the intention to build this properly rather than quickly. If you want to be one of the first businesses on it, the useful thing you can do is tell us what you'd point an agent at. That's what decides what gets built first.
Join the waitlistQuestions people ask
- Can I buy this today?
- No. It's in development, there's no signup, and we won't take money for something that isn't ready. The waitlist is the only thing open.
- Where does the agent actually live?
- On your site, under your brand, as an embedded widget or an iframe. The reasoning and memory run on our infrastructure, which is why you don't have to host anything.
- Who owns the conversations?
- The business the agent belongs to. We process them to run the agent; we don't treat them as ours.
- Isn't this the opposite of "own it, don't rent it"?
- It's a different model, and we're not going to pretend otherwise. When we build a business system for a client, that client owns it outright. Agents is a hosted product — you pay monthly for something we run. Both are honest. They're just not the same offer.
- What if I'd rather have an agent inside a system I own?
- That's the other side of what we do, and it's a real option. Tell us which you'd prefer and we'll say which one actually fits.
- Do you have a launch date?
- Not one we're willing to publish. When we're confident, it'll be on this page.
Next step