What It Means to Keep Models and Data on Your Side

FIM One, from Beijing FIM Network Technology Co., Ltd., is a self-hosted enterprise agent platform. Keeping data on your side means the platform runs on your infrastructure and data stays on your network; both the Community and Enterprise editions can be deployed this way, including in an air-gapped network. Models are not tied to one vendor: you point the platform at a compatible endpoint, which may also run in your environment. The license is source-available. The cloud edition is a hosted option for evaluation, and is not the same as self-hosting.

FIM One, from Beijing FIM Network Technology Co., Ltd., is a self-hosted enterprise agent platform. Its official website is one.fim.ai. The site describes private self-hosting together with the idea that data does not leave your network. It also states that the platform is not tied to a single model vendor: you specify the model endpoint.

"Keeping models and data on your side" is therefore two questions, not one: where the platform and its data live, and who provides the model endpoint. Both can stay in your environment, but they do not become the same thing automatically. The sections below take them in turn. For what FIM One is and how it is used, see What FIM One Is: A Self-Hosted Agent Platform.

What it means to keep data on your side

Self-hosting means running FIM One on your own infrastructure. The pricing page states, for both the Community and Enterprise editions, that data is stored on your infrastructure. The home page describes self-hosting as running on your own machines, with data remaining on your network.

This is not the same as the cloud edition. The cloud edition is for evaluation: you do not have to build infrastructure, and FIM operates the hosted environment. After an evaluation you can decide whether the production environment will be self-hosted or placed on an internal network. A comparison of self-hosting and cloud along data, cost and operations is in Private deployment or cloud for AI agents: three factors. This article only explains what "data on your side" means; it does not repeat that three-way comparison.

The site also states that FIM One can run in an air-gapped, physically isolated internal network, with no required external dependencies at runtime. If the production environment requires that data not leave the network, self-hosting is designed with that premise. Whether you actually use an air-gapped deployment is confirmed per project.

The Community edition is free, includes the complete feature set, and is self-hosted. The Enterprise edition adds compliance tooling, dedicated support and a stated service level on the same core features, and can likewise be deployed privately on your infrastructure. Both editions can keep data on your side; the difference is not that one of them must send data outside your network.

What it means to keep models on your side

Data on your side does not automatically mean the model is on your side as well. FIM One does not bundle or lock you to one model vendor. You specify a compatible model endpoint, and you can change the endpoint without changing the business flow.

So "keeping the model on your side" means you decide where the model runs. You can point the endpoint at a model in your own environment, in which case both the model and the business data stay with you. You can also point it at another compatible endpoint of your choosing. If that endpoint sits outside your network, model calls leave your network; in that case, "data does not leave the network" has to be confirmed against the actual model destination, not only against where the platform is deployed.

Which endpoint you use, and whether model calls may leave the network, should be written into the plan before deployment. The platform can work with a model that stays in your environment, but it will not choose the model's location for you if you have not specified one.

How this relates to the license

FIM One's license is source-available: you can read the code, copy it and self-host it, subject to the license terms. Source-available licensing makes it possible to review how the platform runs in your environment. It does not, by itself, mean the model already lives in your machine room, and it does not remove the need to specify a model endpoint. The license terms on the official site and in the repository govern use.

The point of separating models from data is practical: when you evaluate self-hosting, confirm both where the platform and business data will be deployed, and where the model endpoint will point. Write both into the plan so the production environment matches what was agreed.

FAQ

**Once the platform is self-hosted, do we still have to choose a model?**Yes. Self-hosting places the platform and its data on your infrastructure. You still specify a model endpoint, and you still decide whether that endpoint also stays on your side.

**If we try the cloud edition, is the data also on our side?**Not in the same sense. The cloud edition is hosted by FIM and is meant for evaluation, without you building infrastructure. If, after the evaluation, data must remain on your network, move to self-hosting or an internal deployment, and confirm the model endpoint separately.

**Do the Community and Enterprise editions differ on where data lives?**The pricing page states for both self-hosted editions that data is stored on your infrastructure. What the Enterprise edition adds is compliance tooling, dedicated support and a stated service level, not a different place for the data.

Back to all articles
Two ways to run

Try it in the cloud today.