Does this run models entirely on my machine?

The model runs on your computer via the Ollama daemon, so no third-party AI provider ever sees your prompts. Messages still travel through idapt's cloud (chat, files, routing) and your history is stored in your workspace: private inference with a cloud control plane, not fully air-gapped.

Local model runs are free: there's no per-token charge because the compute is yours. You only pay for cloud models when you choose to use them.

Which models can I run locally?

A curated catalogue of vetted open models (Llama, Qwen, Gemma, Mistral, Phi, DeepSeek distills and more) with quantization options sized to your RAM and GPU.

Which platforms are supported?

Linux and macOS are fully supported. Windows works but is currently experimental.

Can I run it on an idapt cloud computer?

Not by default. A model running on our hardware gives you no privacy benefit and costs more than routing through idapt, so cloud computers ship with Private AI off. You can enable it per machine from the computer itself if you want it anyway, and GPU cloud computers come with it on because that is what they are for.

Compute

Private AI: open models on your own computer.

Run open models on your own hardware with Ollama: private, free, no rate limits.

Start free trial See pricing

What you get

Your hardware, zero per-token cost

Models run on a computer you own, through the idapt daemon: no third-party AI provider involved, free, and never rate-limited.

One-click install

Pick a model on its details page and install it on a connected computer. Hardware-fit hints show what runs on GPU vs CPU.

Local + cloud, one chat

Local models sit beside 200+ cloud models in the same selector. Switch per message; the run tab shows which computer answered.

Use local only

Force a model to run locally, globally or per model: requests fail with a clear message instead of silently using the cloud.

Manage from anywhere

Install, update, and remove models from the web app; progress streams live and loaded models report back automatically.

Your machines, not ours

Private AI is for hardware you control. idapt-hosted computers ship with it off, because a model on our servers is not private; you can still turn it on per machine from the computer itself.

Who it's for

Privacy-conscious users

Keep sensitive prompts away from AI providers: the model runs on your own hardware, while idapt syncs your workspace as usual.

Heavy users

Offload everyday open-model chats to a workstation you already own and pay nothing per token.

Tinkerers

Run the latest open weights on your GPU box and use them with idapt agents, files, and tools.

How it works

1
Pair a computer
Install the idapt daemon on a machine you own and pair it: your laptop, workstation, or GPU box.
2
Install a model
Open any self-hostable model and install it on your computer in a click. idapt manages Ollama for you.
3
Chat locally
Pick the model in chat. It runs on your hardware: free, no rate limits, no third-party AI provider.
4
Tune routing
Optionally set 'use local only' globally or per model for a hard guarantee: no silent cloud routing.

Frequently asked

Ready to try it?

Try 200+ models free for 7 days, then pick the plan that fits. Cancel anytime.

Try now Read the docs

Read the docs

Pair with

What you get

Your hardware, zero per-token cost

Models run on a computer you own, through the idapt daemon: no third-party AI provider involved, free, and never rate-limited.

One-click install

Pick a model on its details page and install it on a connected computer. Hardware-fit hints show what runs on GPU vs CPU.

Local + cloud, one chat

Local models sit beside 200+ cloud models in the same selector. Switch per message; the run tab shows which computer answered.

Use local only

Force a model to run locally, globally or per model: requests fail with a clear message instead of silently using the cloud.

Manage from anywhere

Install, update, and remove models from the web app; progress streams live and loaded models report back automatically.

Your machines, not ours

Private AI is for hardware you control. idapt-hosted computers ship with it off, because a model on our servers is not private; you can still turn it on per machine from the computer itself.

Who it's for

Privacy-conscious users

Keep sensitive prompts away from AI providers: the model runs on your own hardware, while idapt syncs your workspace as usual.

Heavy users

Offload everyday open-model chats to a workstation you already own and pay nothing per token.

Tinkerers

Run the latest open weights on your GPU box and use them with idapt agents, files, and tools.

How it works

Pair a computer

Install the idapt daemon on a machine you own and pair it: your laptop, workstation, or GPU box.

Install a model

Open any self-hostable model and install it on your computer in a click. idapt manages Ollama for you.

Chat locally

Pick the model in chat. It runs on your hardware: free, no rate limits, no third-party AI provider.

Tune routing

Optionally set 'use local only' globally or per model for a hard guarantee: no silent cloud routing.

Private AI: open models on your own computer.

What you get

Who it's for

How it works

Frequently asked

Does this run models entirely on my machine?

What does it cost?

Which models can I run locally?

Which platforms are supported?

Can I run it on an idapt cloud computer?

Ready to try it?

Read the docs

Pair with

Private AI: open models on your own computer.

What you get

Who it's for

How it works

Frequently asked

Does this run models entirely on my machine?

What does it cost?

Which models can I run locally?

Which platforms are supported?

Can I run it on an idapt cloud computer?

Ready to try it?

Read the docs

Pair with