Get started with Visiyon.

Two ways to run Visiyon

There are two ways to get access to a Visiyon platform. The first is our hosted cloud version, available at ai.visiyon.com — sign up, verify your email, and you're working within a couple of minutes. The second is self-hosting the open-source core on your own infrastructure, which gives you full control over data storage, network access, and model configuration.

Self-hosting: requirements

The open-source core is distributed as a containerized application. To self-host, you'll typically need:

• A server or VPS capable of running Docker (2GB RAM minimum for light usage; more if you're running local models rather than calling external APIs).
• A reverse proxy (nginx, Caddy, or similar) if you want to serve the platform over HTTPS on a custom domain.
• API keys for whichever model providers you want to connect — OpenAI, Anthropic, or a self-hosted inference backend such as Ollama or vLLM.

Once running, the platform is configured through an admin panel rather than by editing config files directly, which covers model connections, user permissions, and branding.

Connecting models

Visiyon platforms don't ship with a single hardcoded model provider. From the admin panel you can add API credentials for hosted providers, or point the platform at a self-hosted inference endpoint on your own network. Multiple providers can be connected simultaneously, and access to each model can be restricted per user group — useful if, for example, only certain teams should have access to a more expensive frontier model.

Access control and SSO

For team and enterprise deployments, Visiyon supports sign-in via OIDC-compatible identity providers, so you can tie access to your existing corporate identity system rather than managing a separate set of credentials. Once connected, users can be organized into groups, and each group can be granted or denied access to specific models, tools, or features.

Tools and function calling

Models running through a Visiyon platform can call built-in tools — such as web search or retrieval over uploaded documents — or custom tools you define yourself. Custom tools are configured as HTTP endpoints: you describe the tool's inputs and outputs, and the model decides when to call it mid-conversation based on what the user is asking. This is how teams typically connect Visiyon to internal systems, such as a ticketing tool or an internal knowledge base.

Retrieval-augmented chat

Users can attach documents directly to a conversation, and the platform will index and retrieve relevant passages to ground the model's responses, with citations back to the source material. This works for individual file uploads as well as larger shared knowledge bases that an admin configures for a whole team.

Where to go next

• Full setup instructions and configuration reference live in the GitHub repository README and wiki.
• Release notes and the current roadmap are tracked in Releases.
• Common questions are answered on the FAQ page.
• For deployment help with a production or enterprise setup, see the Enterprise section.