Platform overview
The concepts behind agent4.io — tenants, agents, spaces, knowledge, and channels — and how they fit together.
The model in one picture
Read it top-down — each level is owned by the one above it. The space is the privacy boundary: a customer's history, documents, and memory live there and never leak across it. Two relationships connect the branches: an agent becomes usable in a space by membership, and the knowledge base is retrieved into conversations.
A tenant is your organization's account: your agents, your knowledge, your API keys, your quota.
An agent is a configured persona — system prompt, playbook, tools, and skills. Agents are defined once at the tenant level and reused everywhere: the web widget, Telegram, WhatsApp, and the API.
A space belongs to an end user and is the privacy boundary. Knowledge, chat history, and long-term memory are scoped to the user-and-space pair and enforced at the database level with row-level security. An agent can only respond in a space where it has been added as a member.
Channels are entry points to the same conversations: the embeddable web widget, Telegram bots, WhatsApp numbers, and the REST API. A customer who starts on your website and continues on WhatsApp keeps the same memory and history.
What makes it different
- Answers are grounded. Agents retrieve from your uploaded documents and answer with citations — they don't freewheel on the open internet.
- Memory compounds. Agents remember each customer across sessions and channels.
- Follow-ups are first-class. Scheduled reminders and recurring outreach are part of the agent, not a bolt-on CRM task.
- Five languages built in. UI and agent conversations work in English, 简体中文, 繁體中文, Español, and Русский out of the box.
Where to go next
Start with the Quickstart, then wire up your first channel in Embed the web widget.