This article is auto-synced from its in-app version in Tai.

A handful of ideas power everything in TAI. Learn these and the whole platform clicks into place.

Read this page when you're getting started and want the mental model — what each term means and why it matters — before diving into the how-to guides.

Agents are your AI workforce. Each one is configured for a specific job — an itinerary checker, a customer care helper, a risk analyzer — and does it on demand, around the clock. You decide what an agent does, which skills it can use, what it knows, and which systems it can reach. Every agent is either a Nezasa built-in, a company agent shared across your team, or a personal agent you own — so you can build and test privately, then release the best ones to everyone. Find agents in the Workbench sidebar under Agents.

Skills are what your agents can actually do. A skill is a reusable capability — check an itinerary against your rules, generate a PDF report, search the web — that you attach to any agent that needs it. Define a skill once and share it across your whole fleet. Find skills under Skills in the sidebar.

Knowledge makes answers right, not generic. Knowledge items give your agents grounded, current content to draw on — help articles, procedures, FAQs, and other documents — so they answer from your company's own material instead of relying on general training. Find knowledge items under Knowledge in the sidebar.

Connectors plug agents into your real data. A connector links an agent to an external system — your booking data, a flight-status API, a database, or another SaaS service — so it works with live information, not stale assumptions. Agents can only query the connectors you assign them, so access stays under your control. Find connectors under Connectors in the sidebar.

Channels meet your users where they are. A channel is how an agent is invoked: an interactive chat, a one-shot report run, the REST API, or an embedded widget via the TAI Chat SDK. You choose which channels each agent supports in its settings, and the same agent can serve them all.

Chats are the fastest way to put an agent to work. They're interactive, multi-turn conversations — ask, refine, follow up. Start one with any agent directly in the Workbench under Chats, or through an embedded chat widget in your own app.

Reports turn a single run into a shareable document. Instead of a conversation, an agent runs once against a set of inputs and produces a persisted HTML or Markdown report you can review, share, and export as PDF. Find generated reports under Reports in the sidebar.