Introduction
Welcome to Cubeo AI. This documentation walks you, step by step, from zero to a working AI Agent that talks to your customers, reads your private docs, and takes actions in your other tools.
Who this is for
Builders, ops leads, and product managers — anyone who wants to ship an AI-driven workflow without writing code. You don't need a machine-learning background. If you've used Zapier, Make, or N8N, you'll feel right at home.
What you can build
- Customer-facing AI Agents that answer from your knowledge base, qualify leads, or guide users through a flow.
- Internal AI Agents that automate your team's repetitive tasks — scrape data, update HubSpot, summarise tickets, push to webhooks.
- AI Teams — multiple AI Agents collaborating on a goal, each specialised, with one supervisor coordinating them.
- Anything that mixes your data (uploaded files, websites, Q&A pairs) with external tools (HubSpot, LinkedIn, your own APIs, MCP servers, custom webhooks).
The four building blocks
Every AI Agent in Cubeo is built from these:
- Instructions — plain English describing how it should behave.
- Knowledge Bases — files, websites, and Q&A pairs your AI Agent reads from. See Knowledge Bases.
- Tools — actions your AI Agent can take: call an API, search a CRM, scrape a page, run a sub-agent. See Tools.
- Integrations — external accounts and MCP servers your AI Agents reuse across the team. See Integrations.
How to read this guide
- New to Cubeo? Start with Getting Started.
- Ready to build? Go to How to Create an AI Agent.
- Want safety controls? Read Human-in-the-Loop.
- Need to monitor your AI Agents? See Dashboard.
Each section is self-contained — jump wherever you need.
Connect once, use everywhere
Connect a tool — HubSpot, an MCP server, your own API — once, and every AI Agent in your team can use it.