LobeChat VPSYour AI chat interface, hosted in France

LobeChat is a web AI chat interface published by LobeHub, a third party. It connects to the APIs of many model providers (OpenAI, Anthropic, Google Gemini, Mistral, DeepSeek…) and to Ollama, and adds agents, plugins, vision, voice and image generation.

No. LobeChat is the interface: you plug in your own API keys, billed by each provider, or an Ollama server you install yourself with root access. Without a GPU, Ollama runs on the CPU: that works for small quantised models, stays slow on large ones, and memory becomes the limiting factor since the model must fit in RAM (plan on at least 8 GB, more for larger models). General Purpose and Performance plans have no GPU; on a High Performance plan, the NVIDIA L4 GPU option (24 GB of VRAM, €0.70 excl. VAT/hour billed by the minute, no commitment) accelerates Ollama. No model is included either way.

In your browser. The deployed version is LobeChat 1.x in client database mode: conversations, agents and settings are saved locally on each user's device, with no sync between devices. Messages sent to a model are processed by the provider of the API you chose. Remember to export your data from Settings → Data Storage.

Once the VPS is created, LobeChat is deployed automatically. Open the app in the One Click Apps section of the console: the HTTPS address and the generated access password (ACCESS_CODE variable) are in the Overview tab. Then add your domain in the Domains tab: the HTTPS certificate is issued automatically.

Two ways. For the whole team, add the key as an app environment variable from the console, for example OPENAI_API_KEY: it stays on the server and is used by anyone who enters the access password. For yourself only, enter it in LobeChat's settings: it is then stored in your browser.

Yes, with the same address and access password. This mode has no user accounts: everyone finds their own conversations in their own browser, and nobody sees anyone else's.

One Click Apps need at least 2 vCPU, 2 GB of RAM and 30 GB of disk, with 1 GB of RAM reserved for the system and the platform. Every plan on this page exceeds them, and you can move to a larger size from the console. If you install Ollama on the same server, take considerably more memory, or a High Performance plan with the GPU option.

The platform backs up app volumes every day, but LobeChat does not store your conversations on the server: they are in your browser, and LobeChat's JSON export is the right way to keep them. For the server itself, take a snapshot or add the VPS backup option (from €1.99 excl. VAT/month).

Every night, the platform compares the installed version with the catalogue's and shows an "Update available" banner when they differ. Nothing is updated without your action. Note that LobeHub now develops a version 2, which requires a server-side database; the app offered here remains LobeChat 1.x in client mode.

LobeChat is free, published by LobeHub under the LobeHub Community License (Apache 2.0 with additional terms: commercial use allowed without modifying the code, commercial licence required to distribute a modified version). Once deployed, it runs on your server with root access: its configuration and use are yours to operate. Our support covers the instance itself: the server, its network, its storage and its availability.