Every chat starts cold
One person spends an hour teaching an agent how your company works. The next person does it all again.
Company edition of Holt · Book a demo
Shared memory for your company's AI agents.
What your team learns stays in the company, and every agent can use it. Run it as a memory layer in your product, as a hosted service, or on your own infrastructure, with or without a cloud model.
Closed source. Holt, the personal edition, is open source.
01Why it exists
Holt gives one person an assistant that remembers. Holt Teams asks what that looks like for a whole company, where ten people use agents and each one starts from zero.
One person spends an hour teaching an agent how your company works. The next person does it all again.
What someone learns about a client, a system or a decision sits in their own notes, and leaves when they do.
In some companies the rule is plain: it does not leave the building. The AI tools everyone else uses are off limits.
02What it does
Holt Teams is a shared memory that your AI agents read from and write to. The point is what you get out of it.
What one person or agent works out is there for the next one, so the team stops paying for the same lesson twice.
The memory belongs to you. What your team learns stays inside your company, not in someone else's product.
A company memory needs rules about what gets shared and who sees it. Holt Teams is built around that, not bolted on after.
03Three ways in
The three offerings go from lightest to most involved. You can start small and move up.
Add shared memory to a product you already have. The lightest option, for teams that want the memory and not a new tool.
I run Holt Teams for you. Good for teams with no IT department who still want a company memory.
It runs on your own infrastructure. The most involved option, built for teams that need everything inside their own walls.
04For regulated buyers
Defence, space and finance teams often cannot send their data to an outside model. Holt Teams can run on models that sit inside your own network, so the work happens where the data already is.
Use models that run on your own hardware instead of a cloud service.
The offline mode is designed for places where outside connections are not allowed.
This mode is part of Holt Teams. The open-source Holt does not include it.
Want to know if it fits your setup? Book a demo and tell me the constraints you work under.
05Open source and company
Holt is the open-source personal edition, with a memory you can see on your own machine. Holt Teams is the closed-source company edition. If you only need an assistant for yourself, start with Holt.
Open source. Any LLM, the skills you choose, a private memory you can read.
Closed source. Shared memory for a whole team, in three ways to run it.
06Questions
It is the company edition of Holt. Holt gives one person an AI assistant with a memory they can read. Holt Teams gives a whole company a shared memory that its AI agents can use, so what one person works out is not lost when the chat window closes.
No. Holt Teams is closed source. Holt, the personal edition, is open source and free.
There are three ways. A memory layer built into your own product, a hosted version that I run for you, or an install on your own infrastructure. The memory layer is the lightest option and the on-premises install is the most involved.
Yes. Holt Teams can run on models that live inside your own walls, with no cloud model in the loop. That mode is part of Holt Teams. The open-source Holt does not offer it.
Teams that already use AI agents and keep losing what those agents learn, and teams in regulated places such as defence, space and finance, where the data cannot go to an outside model.
It depends on which of the three options you pick. I talk through that on the demo, not on this page.
I read every request myself and reply by email. You get a short call where I show you what Holt Teams does and listen to how your team works today.
NextRelated
The personal agent OS this grew out of.
A scheduler your AI agent can use, with a person approving every post.
A self-hostable RAG build for teams that cannot use a cloud API.
GoBook a demo
Tell me a little about your team. I read every request myself and reply by email to set up a short call.