TLDR: You have three choices before December 11: rebuild inside OpenAI's new plugin system on OpenAI's timeline, give the method away as files and lose control of it, or move it to a Skill behind a connector you control. TCOM is the third option. Your method sits on our server. Your subscribers connect once from an assistant they already use. You decide who gets in, you see who is using it, and you can remove anyone the day they cancel. In IP mode the method never reaches the subscriber or their assistant. The business keeps running whichever assistant wins next year.
Key Takeaways
- The retirement forces a choice between rebuilding on a vendor's terms, giving the method away, or owning the door.
- TCOM is built for creators who want to own the door: per-subscriber access checked on every call, one-click removal, and a method you can keep server-side.
- Password Protection and IP Protection are two different promises; pick per Skill.
- Subscribers connect once and reach every Skill you grant them, grouped in Stacks.
- Your dashboard shows who is in and what they are using.
- Nothing about this depends on OpenAI's calendar.
The three doors
When a platform retires the thing your product runs on, you can:
- Rebuild inside the vendor's new system. For GPTs that means plugins, a private-by-default start, and a separate submission to go public. Fine if OpenAI's roadmap is your roadmap.
- Send the method as files. Fast, and over. Once a document is on someone's machine it is on everyone's machine, and your subscription is a suggestion.
- Put the method behind a door you control and let a supported assistant knock. That is what a Skill on TCOM is.
What moves, what stays
What moves: the instructions, the knowledge files, the way you want the work done. That becomes a Skill.
What stays: your subscribers, your pricing, your brand. You add the subscribers to your Stack. They connect from Claude or Perplexity, or from the next assistant we verify, and your Skill shows up as a tool they can call.
Two kinds of protection, chosen per Skill
Password Protection means the subscriber must be on your list before the assistant receives the Skill, and the check happens every time they call it. It is the right choice for a Skill whose value is in the outcome and whose text you do not mind the assistant having for the chat.
IP Protection means the subscriber's assistant sends the request, our server runs your method with the model provider's API, and only the result goes back. The method never reaches the subscriber or their assistant. This is for the Skill that is your actual secret. It runs on credits because our server does the work.
The part a GPT on its own never gave you
A GPT on its own could not tell you who was using it, and it could not stop a subscriber who cancelled. A Stack can. The creator dashboard lists your subscribers and what they reach; removing one takes one click, and their next call is refused. On the larger plans you also get webhook automation, so your billing system can add and remove people without you touching the dashboard.
What TCOM is not
It is not a new chat app your subscribers have to learn. It is not a marketplace that takes a cut of every conversation. It is not a GPT clone that will be retired next. It is a server that holds your Skills and a connector that a supported assistant can use, with you holding the list.
How to start
Connect free, try the two demo Skills from Claude, and watch the Password one hand over its text while the IP one hands over only a result. When that makes sense, pick a plan, put your first Skill in, and add your real subscribers.
