Skip to main content

Core Concepts

Projects​

A project is the top-level container for everything on Vois AI. All agents, skills, and integrations belong to a project. Use separate projects to isolate environments (e.g. acme-production vs acme-staging) or different products.

Agents​

An agent is an AI-powered voice or text bot. Each agent has:

  • A model — the underlying LLM (GPT-4o, Gemini Flash, etc.)
  • A system prompt — instructions that define its role and behavior
  • A language — the language it speaks and understands
  • Attached skills — reusable behaviors
  • Attached tools — integrations it can call at runtime

Skills​

Skills are reusable instruction blocks. A skill might describe how to greet a caller, how to escalate to a human, or how to respond formally. You write the instructions once and attach the same skill to many agents.

Skills can reference tools using {{tool:tool_name}} syntax in their instructions.

Integrations​

Integrations connect agents to external systems. Each integration exposes one or more actions — callable tools the agent can invoke during a conversation (e.g. ocr_read_image, lookup_customer).

Integrations are installed per-project and are then available to all agents in that project.

Playground​

The playground lets you test an agent with real voice or text input before deploying it. You can attach files (images, PDFs) and test OCR or document workflows end-to-end.