Why Seamcult has no chat of its own

One of the first things people ask when they open Seamcult: where do I type? There is a canvas, a mannequin, a review tool — and no chat window. That is not an oversight. It is a decision, and we would like to explain it, because it shapes what Seamcult is. 💬

Chat is somebody else's whole product

Claude, ChatGPT, Cursor and the rest are not "a chat with a model". They are companies whose entire face is the chat: conversations that branch and come back, history you can search and archive, the same thread open on your laptop and your phone, a picker for the model you trust today, files, voice, memory of what you told it last month. Hundreds of people polish that experience every day, because it is the product. 🏗️

We could build a text box in the corner of a pattern editor and call it a chat. Next to the one you already use, it would be a toy — and you would feel it in the first minute. Our users are pattern makers, not people who want a second, worse version of a tool they already like. Handing them a homemade chat would be like handing a tailor a plastic pair of scissors because we happened to have one. ✂️

So we did the opposite. Seamcult is an MCP server: your agent comes to the canvas, with the chat you already know, the model you already chose and the history you already have. We bring the drafting table — the canvas, the mannequin, the drafting rules and the auditors. The conversation stays where it is good. 🤝

One subscription, not a dozen

The second reason is money — yours. A built-in chat means we would be paying for every token your agent spends, and we would have to bill you for it. That is a second subscription for the same model you are already paying for, just because it is being used in a different window. 💸

Many services do exactly this, and we understand why: it is the obvious way to charge. But look at it from your side. You subscribe to one assistant. Then your pattern tool wants its own subscription to it, your spreadsheet wants another, your mail client a fourth. A dozen subscriptions to one model, in a dozen places, each with its own quota and its own bill. Nobody has that budget, and nobody has that patience. We honestly don't know whether this model works even for the services that try it — and we would rather not find out by building our whole business on it. 🧾

There is a catch in the price of tokens, too. A chat of our own would buy them through the API, at API rates. Your own subscription buys the same tokens far cheaper: a monthly plan with weekly limits is a bulk deal, and the agent you already use is running on it. We have been watching how many tokens an agent actually spends on a pattern — drafting, reading its work back, reading the audits and your notes — and at API rates, even on the cheapest models, one pattern comes out at around $20. 🧮 In a market with plenty of tools competing for the same sewist, that is not a price, it is a reason to leave. Through your subscription, the same pattern costs you a slice of a plan you are paying for anyway.

It would also lock us into a pricing model before we know what Seamcult is worth, and to whom. We wrote about that here: while we are in early access we are still learning who uses the tool and for what. Charging for tokens would answer that question for us, and probably wrongly. 🌱

What you get instead

Not having a chat turned out to be a feature. 🎁

• The model is yours to choose. A new one came out this morning? Point your agent at it. We never have to catch up.
• Your history stays with your agent. Every drafting conversation sits next to all your other work, searchable, synced to your other devices, archived when you are done.
• You can change your mind. Start a project with one agent, continue with another. The project lives in Seamcult; the agents come and go.
• We get to focus. Every hour we don't spend on chat UI goes into geometry: the mannequin, the audits, the documents. That is the part nobody else is going to build for you.

In practice

Connect any agent that speaks MCP — from Claude Code and Cursor to ChatGPT and Copilot Studio — and talk to it where you always do. On the canvas you watch the pieces appear, and when something is off you circle it and write a note, or draw right over the piece. That note goes to the agent as part of the conversation it is already having with you. Review is the one thing you do inside Seamcult; everything else you say, you say in your own chat. ✍️

Will it always be this way? We don't promise never. If one day a chat inside Seamcult would make your work better rather than merely more familiar, we will build it — and you will read about it here first. Until then: bring your agent, and let's draft. 🪁