What it is

Zola is an open-source TypeScript chat interface for multiple AI model providers. It lives in the AI interaction and interface ecosystem and gives users one chat surface for OpenAI, Mistral, Claude, Gemini, and local Ollama models. The project uses Apache-2.0 and is built with Next.js, prompt-kit, shadcn/ui, Tailwind CSS, Vercel AI SDK, and Supabase.

The problem it addresses is provider fragmentation. Users who need cloud and local models often switch between vendor websites, local clients, and API wrappers. Zola tries to reduce that friction with one responsive UI, prompt workflow, and deployment path. It also lists bring-your-own-key support through OpenRouter.

Key capabilities

• Multi-model support covers OpenAI, Mistral, Claude, Gemini, and Ollama, including remote APIs and local inference.
• Bring-your-own-key access through OpenRouter is listed as a supported way to connect model providers.
• File uploads are listed as a feature, while INSTALL.md covers unlocking auth and file uploads.
• The UI supports user system prompts, multiple layout options, responsive behavior, and light and dark themes.
• Ollama support includes automatic model detection, allowing Zola to find locally pulled models without manual listing.
• The stack uses prompt-kit, shadcn/ui, Tailwind CSS, Vercel AI SDK, and Supabase for components, styling, model integration, auth, and storage.
• Full MCP support is listed as work in progress, not as a finished feature.

Who uses it and how

• Users can clone the repository, run npm install, set an OpenAI key in .env.local, and start the development server.
• Users wanting local inference can install Ollama, pull a model such as llama3.2, and run Zola with npm, relying on automatic model detection.
• Self-hosters can use docker-compose.ollama.yml to bring up Zola with Ollama in one local workflow.
• Users can use the Vercel deployment button, while auth and file uploads require additional INSTALL.md configuration.
• Users can evaluate direct provider access, OpenRouter-based bring-your-own-key access, and local Ollama use from the same chat surface.

Getting started

Typical installation uses git clone, npm install, and npm run dev, with an OpenAI key in .env.local for cloud models. Docker is available through docker-compose -f docker-compose.ollama.yml up, and a hosted option uses the Vercel deployment button.

When to use it — and when not to

Zola fits users who want an open-source, self-hostable chat interface for several model providers and local Ollama models. It is less suitable for someone who wants a finished hosted product with no operational work, because the README describes a beta release with an evolving codebase and points to INSTALL.md for auth and file uploads. Self-hosters must manage provider keys or OpenRouter access, plus Supabase-based auth and storage setup where needed.

git clone https://github.com/ibelick/zola.git
cd zola
npm install
echo "OPENAI_API_KEY=your-key" > .env.local
npm run dev

# Install and start Ollama
curl -fsSL https://ollama.ai/install.sh | sh
ollama pull llama3.2  # or any model you prefer

# Clone and run Zola
git clone https://github.com/ibelick/zola.git
cd zola
npm install
npm run dev

git clone https://github.com/ibelick/zola.git
cd zola
docker-compose -f docker-compose.ollama.yml up