Build your own private ChatGPT for companies with a free, open-source stack
A ChatGPT the company owns: hosted for them, connected to their email, Drive and Slack, and sold per seat for less than ChatGPT Team.
Companies pay about $25 a seat a month for ChatGPT. A private version with the same chat, their own data and their own controls can be sold for $12 a seat.
The stack
- LibreChat ↗The full ChatGPT-style app: chats, files, users, agents, adminMIT
- Claude ↗The brain: reasons, plans, writes the code and talks to usersPaid API
- MCP connectors ↗Open-protocol connectors that let the chat read email, Drive and SlackMIT
How to build it
- 1Deploy LibreChat on the client's cloud or yours, behind their single sign-on.
- 2Plug in Claude as the model, with a spend cap per user.
- 3Add MCP connectors for the tools the team already lives in.
- 4Set up agents for the three jobs people ask for most.
- 5Sell it per seat, with admin reports on usage and cost.
One client with 800 seats at $12 a seat is $9,600 a month.
- Law firms
- Accounting firms
- Hospitals
- Schools and colleges
- Government contractors
- AI image generator API →
- Qwen-Imagefree
- Claudepaid
- A rented cloud GPUpaid
- AI employee for one niche →
- Claudepaid
- OpenClawfree
- Lightpandafree
- AI receptionist →
- Chatterboxfree
- Whisperfree
- LiveKitfree
- Claudepaid
Common
questions.
Straight answers. If yours isn't here, ask on a 20-minute call.
What is this list?+
Every week our reels break down an AI product a real company is making money with, and the free, open-source stack you could build your own version on. This page keeps all of them in one place, newest first, with the tools, the licences, the money maths and the reel.
Is the stack really free?+
The software is. Every tool listed is open source or fair-code, and the licence is shown next to it so you can check commercial use. You still pay for hosting, GPUs and the AI model's API, which for most of these products is a few hundred dollars a month at launch.
Which licences allow commercial use?+
MIT and Apache 2.0 allow commercial use with attribution. AGPL-3.0 does too, but if you change the code and offer it as a service you must publish your changes. Fair-code licences such as n8n's Sustainable Use License allow internal and client use but restrict reselling the tool itself, so read them before you build on them.
Can Axionry build one of these for me?+
Yes. We build AI products on open-source stacks with $0 upfront: each checkpoint is invoiced only after you have seen it working. Book a 20-minute call and bring the niche you want to sell to.
Want your own private ChatGPT for companies built for your niche?
Twenty minutes, straight to the engineer. No sales rep, no deck.