A New Way To Run A Business

Make your business
agent-native.

Sell digital products, ship physical goods, capture leads — from one platform that also runs your site, your customers and your calendar. All of it behind an open API, so the AI you already use can work it directly.

Start for Free See what an agent did with it

No credit card required. Connect Stripe when you're ready to charge.

Revenue

Three ways to get paid, built in

Most businesses run at least two of these. Running them in separate tools is why nobody can tell you which article sold the thing. Here they share one customer record and one checkout.

📥

Monetize digital content

Put a price on an article, a report, a dataset, a course. Sell it outright, gate it behind a membership tier, or give the first half away and charge for the rest — the paywall cuts exactly where you put it, and the free part still gets indexed.

📦

Sell physical products

A real catalog: shipping addresses, inventory counts, order status, fulfilment. Buyers filter by the attributes your products actually vary by, and every order lands against the customer who placed it.

📋

Generate leads for services

Not selling a SKU? Put a form on any page and point it at a list. It creates the contact, joins the list, logs the enquiry, and records which page produced them — so your pipeline knows where it came from without a tracking spreadsheet.

💳

Your Stripe, your money

Connect your own Stripe account in a click. One-time payments, subscriptions, renewals and the customer portal are handled, and funds settle to you directly — we never hold your revenue.

One System

The whole operation, under one roof

The old stack was five products and a pile of integrations holding them together — which is fine when a person is the glue, and useless when you want an agent to be. Here it is one system, so what brought a customer in is attached to the order they placed.

📝

Publishing

A block editor for pages and long-form articles, a media library with responsive images, and public pages that render server-side with clean URLs and structured data so they can actually be found.

🛍

Storefront

Product pages, a filterable catalog, checkout, digital delivery, order management and a customer portal — on your own account, at your own URLs.

👤

Customers

Contacts, segmented lists, enquiry history, purchase history and internal notes. Sales opens a record and sees what the person read and what they bought, not a CSV row.

📈

Measurement

A/B test a headline, a price or a form on any block. Conversions count against the same customer records everything else uses, so a result is a revenue number rather than a click number.

Bring Your Own Agent

We don’t sell you the AI. We make your business reachable by it.

There is no bundled assistant here, and no model you’re locked into. The whole platform is an open API and a command-line tool, which is what the agents you already pay for are good at driving — so they can publish the page, price the product, answer from your own documents and work the follow-up list, end to end.

# publish the launch article, written as markdown
modelpad notes create --title "The 2026 Buyer's Guide" --markdown guide.md

# put it on the storefront and price it
modelpad products create --name "Buyer's Guide (PDF)" --set price=29.00

# find what it already knows before answering a customer
modelpad notes list --query search="do you ship to the EU"

# work the follow-up list — who enquired, and what did they ask
modelpad contacts list --query list=pricing-enquiries
modelpad contacts interactions <id>
🤖

Whatever agent you already use

Claude Code, Codex, Cursor, a harness you wrote yourself. Anything that can run a command works on day one, and an MCP endpoint exposes 46 tools over streamable HTTP for clients that prefer tool calls to a shell.

🌐

Any model, including your own

Nothing here is tied to one vendor. The in-app assistance runs on hosted Claude, on OpenAI, or on anything you host yourself — Ollama or any OpenAI-compatible endpoint — so your data can stay on your hardware if that matters.

Built for headless runs

One env var authenticates. JSON to stdout, human text to stderr, so piping is safe. Markdown converts both directions, so a model never learns an internal format. Updates read-then-merge, so setting one field doesn’t blank the record.

🔑

A key per agent, revocable

Issue a separate credential for each agent or automated run, see them all in one list, and revoke any of them independently. An autonomous run gets its own blast radius.

Case Study

One instruction. A whole campaign.

A real run with an off-the-shelf coding agent — no custom integration, nothing bundled from us. The kind of work that used to mean a prospecting tool, a scheduler, a website builder and an afternoon of copy-and-paste between them.

The instruction “Find me prospects worth talking to, get calls on the calendar, and build a landing page aimed at that specific group.”
🔍
It researched the prospects — and filed them
Not a spreadsheet it hands back to you. The agent created each one as a real contact record and grouped the cohort into its own list, in the same CRM the rest of the business already runs on.
📅
It put the calls on the calendar
Each call created as an event linked to the contact it belongs to. The person and the meeting are one record, so nobody is cross-referencing two tools to work out who Tuesday at three is.
📄
It built the page for that exact cohort
A landing page written for those prospects specifically — their industry, their objections — assembled from blocks and published live, with a form on it pointing back at the very list it had just built.
Three products’ worth of work, one instruction. Not because the agent was special — it was a stock coding agent with an API key — but because the contacts, the calendar and the site were the same system. Wire those together with integrations and an agent has three APIs, three auth flows and no idea that the contact and the invitee are the same person. Here it just had the keys.
Under The Hood

Why the agent is any good at it

An assistant is only as useful as what it can see. Everything here is one connected graph with search built in, which is the difference between an agent that guesses and an agent that knows.

🔗

Everything is connected

Customers, orders, products, pages and documents are linked by typed relationships — Purchased, Interested In, Lead From, Featured On. "Who bought this and what did they read first" is one question, not a data project.

🧮

Search that finds the right thing

Keyword matching catches the SKU and the order number. Vector similarity catches the customer who phrased it their own way. Both run together and return one result set, so an agent answering support gets the actual document.

Always current

Documents are indexed the moment they are saved, in the same write. No nightly job, and no window where your agent is answering from last week’s prices.

🎯

Grounded, not improvised

Point retrieval at one product line or one folder and the answers stay inside it. Your agent cites your material instead of inventing a policy you never wrote.

Access Control

Free, gated, or paid — per item

Set it on the thing itself, not with a plugin. Leave the top of the funnel open so it gets found, and make everything past it earn an email address or a payment. One item can offer more than one way in.

🌐

Public

Open to everyone and indexable. Where your SEO work pays off.

🔒

Registration required

Free, but you get the contact. Right for benchmarks and templates.

💳

One-time purchase

Pay once, keep access. Reports, datasets, training material.

💎

Subscription tier

Recurring access to a body of content. Predictable revenue.

The Rest Of It

The unglamorous parts, already done

The work between a good idea and a business that runs. None of it is exciting, all of it is required.

🌐

Public pages that render server-side

Documents and products publish as real HTML with canonical URLs, structured data and a sitemap that updates as you go — so the corpus you built for your agents is also indexable by everyone else’s.

An editor humans want to use

A block editor with inline AI assistance, cloud or local models, and personas per document. Agents and people write into the same store, in the same format.

🔨

A documented REST API

The web app is a client of it, so nothing is held back for the UI. Predictable resources, consistent pagination, JSON errors, and a reference with a runnable example on every endpoint.

📤

No lock-in by obscurity

Documents, relationships, contacts and orders read back out in the same shape they went in. Your graph is yours, and it leaves as cleanly as it arrived.

Read the API documentation

Make it reachable. Then point your agent at it.

Create an account, put your business in it, and hand whichever AI you already use an API key. Connect Stripe whenever you’re ready to start charging.

Create Free Account