Build & Resell
Resell AI Agents Without Running the Infrastructure
Package AI agent instances into your own product, at your own price, under your own brand. You keep the customer relationship. We run the compute.
The Idea
Plenty of people can sell an AI assistant to a market they already understand — an agency with fifty clients, a SaaS with an audience, a consultant whose customers keep asking for the same automation. What stops them is everything underneath: containers, scaling, SSL, uptime, model routing, and the pager at 3am.
Build & Resell is that underneath, as an API. You deploy agent instances programmatically, wrap them in whatever product and pricing you like, and sell to your own customers. We never contact them, and we never appear in your product.
What You Get
A REST API
Create an agent instance, send it input, stream the reply back over SSE, and stop, restart or delete it — all from your own code with your own API key. Instances come out of a warm pool, so a deploy is ready in seconds rather than minutes.
A storefront template
A working Next.js storefront you can deploy as-is or fork: sign-up, checkout, and a customer dashboard already wired to the API. It is a starting point for your product, not a hosted page you rent from us.
Webhooks
Register an HTTPS endpoint and your storefront is told when one of the bots you sold changes state, when your prepaid balance runs low, and if your account is suspended. Deliveries are signed, retried on failure, and carry an id you can deduplicate on.
A roster you control
Every bot you deployed through the API, listed in one place, with stop, restart and delete. Deleting your own storefront never touches the bots your customers depend on — those stay yours to manage until you decide otherwise.
How the Money Works
Billing is prepaid and deliberately boring. You top up a balance; deploys draw it down. There is no monthly platform fee, no minimum commitment, and unused balance is refundable.
$5 per Lite deploy, $10 per Pro deploy. The fee includes the bot's first 30 days of hosting. After that, hosting meters per running hour — $0.007/hour Lite, $0.014/hour Pro — which works out to roughly $5 and $10 over a 30-day month for a bot that runs continuously. A stopped or sleeping bot accrues nothing.
Models are funded by you. Each resold bot runs either on the free Agnes model, or on your own OpenRouter account through a per-bot key we mint with a spend cap, so one customer's bot cannot drain your balance. Model cost sits with you, which is why the deploy fee covers infrastructure only.
Balance, pending usage, and the full ledger of top-ups, deploy fees and hosting charges are all readable from the API, so your own billing can reconcile against ours instead of trusting a dashboard.
Your Customers Stay Yours
One design decision is worth stating plainly, because it is the one most likely to matter to you later. Deleting a storefront revokes its API keys and takes the shopfront down. It does not delete the bots you sold through it.
Those bots are live assistants belonging to people who were not part of that decision, and tearing them down as a side effect of an unrelated cleanup would be the wrong default. They stay on your roster, where you can stop or delete each one deliberately.
What You Can See, and What to Tell Your Customers
You are the controller of your customers’ data and we are the processor. That is not a formality — it means the duty to tell them what you can reach is yours, so here is exactly what that is.
By default the export endpoint returns how the bot is set up — its config plus the persona files you wrote — and nothing about who talked to it. Three categories go further and have to be asked for by name: sessions is what your customer actually said, knowledge is the workspace the assistant built from those conversations, and skills can quote them too, because assistants write skills of their own accord from what was discussed. Any export containing those is recorded separately in our audit log. On Hermes bots a sessions export is partial rather than complete: live conversations sit in SQLite, which we do not yet export safely.
So conversations are reachable — deliberately, because you are the controller and may have to answer your customer’s own access request. What they are not is casual: they are never in a routine backup, never on your storefront roster, and never retrievable without leaving a record.
Channel credentials are the exception that genuinely never leaves. On OpenClaw, asking for credentials returns an inventory — which channels have credential files, how many profiles and bytes — and never the keys themselves. Read it as “this bot has WhatsApp set up”, not as proof the pairing is currently live, since the files exist before a scan is confirmed. Hermes rejects the category outright. Either way the keys stay put, because a WhatsApp session authenticates your customer rather than their bot and cannot be moved to another host anyway — you re-pair on the destination.
One thing runs the other way, and is worth reading twice. App integrations connect through your Composio key, so a customer who connects Gmail is granting that access into your workspace, where you can view, use or revoke it from Composio’s own console — outside anything we control. We cannot contain it, so we disclose it: on OpenClaw bots the assistant is instructed to tell your customer who operates it before they approve, naming your storefront domain when you have one. That is an instruction to a language model rather than a hard gate, and Hermes bots have no in-chat connect flow at all. Treat it as a floor, not a substitute for saying so in your own terms.
Limits
- 25 live instances at once on a Scale account
- 30 creates per minute and 300 agent responses per minute
- 2 vCPU, 4 GB RAM, 40 GB disk per instance
- 5 webhook endpoints per account
These are launch defaults chosen to be safe rather than to cap the business. If your storefront outgrows them, tell us and we will raise them.
Getting Started
Build & Resell is open to everyone — there is no application and no waiting list. Sign in, open Build & Resell in the dashboard, add a card and top up a balance, then create an API key and deploy your first instance. Nothing is charged until you top up, and the balance you load stays refundable.
If you are planning something the launch limits above do not fit, email [email protected] and tell us what you are building.
Questions
What does Build & Resell actually give me?
It is a resell layer over the same managed fleet that runs the OpenClaw Launch dashboard. You get a REST API for creating and driving AI agent instances, a Next.js storefront template with sign-up and checkout already wired up, signed webhooks so your own systems learn about state changes, and a roster of every bot you deployed. You package and price the product; we run the compute, the scaling and the operations underneath it. Your customers are yours — we never contact them and never appear in your product.
How much does it cost?
Billing is prepaid: you top up a balance and deploys draw it down. A Lite deploy costs $5 and a Pro deploy costs $10, and that fee includes the bot’s first 30 days of hosting. After 30 days, hosting meters per running hour — $0.007/hour for Lite and $0.014/hour for Pro, which is about $5 and $10 respectively over a 30-day month if the bot runs continuously. A bot that is stopped or asleep accrues nothing. There is no monthly platform fee and no minimum, and your unused balance is refundable.
Who pays for the AI models?
You do, and you choose how. Every resold bot can run on the free Agnes model at no model cost to you, or on your own OpenRouter account — in which case we mint a per-bot runtime key from your provisioning key with a spend cap on it, so a single customer’s bot cannot drain your account. Because model funding sits with you rather than with us, the per-deploy fee covers infrastructure only.
How many bots can I run?
A Scale account allows 25 live instances at once, with a rate of 30 creates per minute and 300 agent responses per minute. Each instance gets 2 vCPU, 4 GB of RAM and 40 GB of disk. If you outgrow those numbers, talk to us — they are launch defaults set to be safe, not a ceiling on the business.
What happens if my balance runs out?
Auto-reload tops the balance up from your card when it falls below a threshold you set, starting at $5 per reload. If a top-up fails and uncollected hosting reaches $2, the account suspends: the bots stop and the model keys are disabled until the balance is settled. You can subscribe to the balance.low webhook to hear about it well before that point, and the balance and its full activity history are readable from the API at any time.
What happens to my customers’ bots if I stop reselling?
They keep running until you say otherwise. Deleting a storefront revokes its API keys and takes the shopfront down, but it deliberately does not tear down the bots your customers are using — those would be live assistants belonging to people who did not make that decision. The bots stay on your roster, where you can stop or delete them individually whenever you choose.
Is it available now?
Yes. Build & Resell is open to everyone — sign in, open Build & Resell in the dashboard, add a card and top up a balance, and you can create your first API key and deploy through the API straight away. There is no application and no waiting list. Nothing is charged until you top up, and the balance you load stays refundable.