BETA

Self-hosted Claude for your team

Your team's private Claude, on your own GPU.

Q: Is OwnLLM self-hosted or SaaS?

Hybrid. We host the control plane in Europe. Models and inference run on your machine. You get SSO, audit, and updates without the DIY pain.

Q: Why is it cheaper than per-seat AI subscriptions?

Per-seat AI gets expensive past 20 to 30 people. OwnLLM is one flat fee. You bring the GPU machine.

Q: Does it replace Claude Code, Cursor, or Copilot?

No. Keep Cursor and Claude Code. Route the cheap, sensitive, or internal stuff through your private API.

Q: What machine do we need to start?

A Mac Mini, Mac Studio, or RTX workstation. Bigger models need more VRAM. We help you pick.

Q: Can we choose our own models?

Yes. Curated open-weight models via Ollama: Llama, Mistral, Qwen, DeepSeek, Phi, Gemma. Swap anytime.

Q: Does tool calling work with every model?

Only on models that support it. If a client sends tools to a chat-only model, you get a clear error.

Install one app on a Mac, PC, or Linux box. Your team gets web chat, an OpenAI-compatible API, and SSO. No data leaves your box.

Book an AI cost audit ROI calculator

How it works

From a GPU machine to a team Claude in 3 steps.

Book an AI cost audit

Install the app

On a Mac, PC, or Linux box you already own.

Paste your pairing key

We auto-configure the tunnel, Ollama, and your models.

Share the URL with your team

Web chat for everyone, API keys for your devs.

One machine per team. No cloud lock-in.

What you get

Everything Claude Team has, running on your machine.

Your data stays on your hardware

Inference runs on your GPU. We never see your prompts.

Login with your company SSO

Magic link, then SAML/OIDC and SCIM when you grow.

Plug Cursor and Claude Code in

An OpenAI-compatible URL. Your devs change one env var.

Web chat for everyone else

A team URL. People log in like Slack and chat.

Flat price. No per-seat sprawl.

One subscription replaces Claude + Copilot + Cursor seats.

Savings & capacity

Pick the right model. Pay flat. Scale when you grow.

Open-weight models curated for your hardware. Swap them anytime.

Flat cost

5-10x

cheaper than per-seat AI

Over 2 years for a 50-person team.

Fast setup

<15m

to first message

Guided pairing and model download.

Scalable

10-50+

users per machine

Mac Mini for small teams. Mac Studio or RTX when you scale.

LlamaMistralQwen CoderDeepSeekPhiGemma

ChatCodeVisionToolsThinking

Tool calling is auto-detected per model. Agentic clients get a clear error if a model can't handle tools.

Security & compliance

Built so your DPO says yes.

EU-hosted control plane. Inference on your machine. DPA included.

Clear story for your DPO

We see metadata for billing and audit. Nothing else by default.

Outbound tunnel only. No open ports.
SSO, SCIM, role-based access.
Hashed API keys with budgets and per-model scopes.
Audit logs separated from message content.
Hosted in Europe. DPA on request.
Prompts not logged by default. Configurable retention.

Pricing beta

One subscription. Your whole team.

Stop paying per seat for Claude, Copilot, and Cursor. One flat fee. One machine. Your team.

Solo

For a single dev who wants a private OpenAI-compatible URL.

5 EUR/ mo

Replaces your OpenAI API key. Up to 3 users.

Up to 3 users
Live
1 paired machine
Live
1 active model
Live
Magic link auth
Live
OpenAI-compatible API
Live

Get started

Team

For 1 to 10 people. Web chat and API.

79 EUR/ mo

Replaces ~10 Claude Team seats

Up to 10 users
Live
1 paired machine
Live
3 active models
Live
Magic link auth
Live
OpenAI-compatible API
Live
Chat web
Beta

Get started

Startup

Recommended

For 10 to 50 people. SSO, audit, all the essentials.

299 EUR/ mo

Replaces ~50 Claude Team seats or stacked Claude + Copilot subscriptions

Up to 50 users
Live
8 active models
Live
SSO SAML / OIDC
Beta
90-day audit logs
Soon
API budgets and scopes
Live
Capability-aware model routing
Live

Get started Not sure yet? Book a free audit

Enterprise

For 80+ people. SCIM, custom domain, priority support.

899 EUR+/ mo

Replaces Claude + Copilot + Cursor Teams stacks at 80 to 200 employees

Users on quote
Ask us
20+ active models
Live
SCIM 2.0
Soon
12-month audit export
Soon
Custom domain
Ask us
4h support and compliance services
Ask us

Talk to sales

Frequently asked questions

The objections your CTO, DPO, and developers will raise.

Is OwnLLM self-hosted or SaaS?

Why is it cheaper than per-seat AI subscriptions?

Does it replace Claude Code, Cursor, or Copilot?

What machine do we need to start?

Can we choose our own models?

Does tool calling work with every model?

Have a specific question? hello@ownllm.app