Skip to main content
AI Agents let Fliqr AI answer questions you did not script in advance. You configure a model, a system prompt, knowledge, and tools. The agent reasons at runtime and replies in the contact’s language — then hands off to a Flow or a human when needed.

​
When to use an AI Agent

Many workspaces combine both: a Flow for intake, an Agent for Q&A, and human handover for edge cases.

​
How an Agent is built

  1. Provider + model — OpenAI, Gemini, Claude, DeepSeek, xAI
  2. System prompt — role, constraints, output format
  3. Knowledge / files — grounded answers from your content
  4. Functions & MCP — live actions and external systems
  5. Fallback — Default Reply Flow, keyword rules, or human inbox
Deep conceptual model: AI Agent concepts.

​
Start here

Connect providers

Bring your own API keys or use included quota.

Create an Agent

Spin up an agent and attach business data.

Prompt engineering

Write prompts that stay on-brand and reliable.

Knowledge sources

Ground answers in URLs, docs, and transcripts.

Functions & actions

Call APIs and trigger Flows from the model.

Human handover

Route to your team when AI should stop.

​
Capabilities map

AI Agent concepts

Components, models, and Agent vs Flow.

Flows

Scripted paths and AI Action blocks.