> ## Documentation Index
> Fetch the complete documentation index at: https://docs.quickblox.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Agent types
> Compare QuickBlox AI Agent types — Knowledge Agent and Workflow Agent — and choose the right one for your use case
QuickBlox AI Agent comes in two types: a **Knowledge Agent** for open-ended AI conversations, and a **Workflow Agent** for structured, designed flows. This page explains how each type works, how to choose between them, and which settings they share.
## Knowledge Agent vs Workflow Agent
| | Knowledge Agent | Workflow Agent |
| ----------------------------- | ------------------------------------------------------ | ---------------------------------------------------------------------------------- |
| **How it answers** | Freely, drawing on its instructions and knowledge base | Follows a designed workflow graph |
| **Conversation structure** | Open-ended — the user drives the conversation | Structured — the agent guides the user through defined steps |
| **Knowledge base** | Yes — used for all answers | Yes — available via the AI response (RAG) node and as a fallback option |
| **Workflows** | Not available | Yes — designed and managed on the Chat Workflows tab |
| **Ideal for** | General Q\&A, support bots, FAQ agents | Lead capture, booking flows, structured support triage, multi-step data collection |
| **Requires advanced feature** | No | Yes |
## Choose an agent type
When you create a new AI Agent, you are asked to choose the agent type.
1. Go to ***Dashboard => YOUR\_APP => AI Agent***.
2. Click **Create**.
3. Under **Agent type**, select **Knowledge Agent** or **Workflow Agent**.
4. Fill in the remaining details and click **Create**.
The agent type cannot be changed after the agent is created. If you need a different type, create a new agent.
## Knowledge Agent
A Knowledge Agent answers freely, drawing on its **Instructions** and its **Knowledge Base**. The user drives the conversation, and the agent responds to each message in turn — there is no predefined path. This is the default, simpler agent type and does not require the advanced feature.
It is a good fit for:
* General Q\&A and FAQ-style assistants
* Support bots that answer from uploaded documentation
* Conversations where flexibility matters more than a fixed sequence of steps
Most of what shapes a Knowledge Agent lives in two places:
* **Instructions** — the prompts that guide tone, behaviour, and how the agent answers.
* **Knowledge Base** — the files, websites, and pages the agent draws on for retrieval-augmented (RAG) answers. See [Knowledge base](/ai-agent/knowledge-base).
You can still collect information up front with an [Intake form](/ai-agent/intake-forms) before the open-ended conversation begins.
## Workflow Agent
A Workflow Agent follows a designed **workflow** — a visual graph of steps that controls exactly what the agent says, what it collects, which systems it calls, and when the conversation ends. This type requires the advanced feature.
Workflows are created and managed on the **Chat Workflows** tab. One agent can have many workflows, but only **one workflow can be active** for live conversations at a time.
To learn how workflows work and how to build one, see:
* [Workflows overview](/ai-agent/workflows/overview) — what a workflow is and when to use one
* [Building a workflow](/ai-agent/workflows/editor) — create, publish, and activate a workflow
Two behaviours are worth knowing when choosing this type:
* **After a workflow ends**, each End node decides what happens next — continue with AI answering, answer from the knowledge base, or close the conversation. See [Workflow nodes](/ai-agent/workflows/nodes#after-workflow-ends).
* **When no workflow is active** — for example, right after creating the agent or after deactivating every workflow — the agent falls back to standard AI answering using its instructions and knowledge base, behaving like a Knowledge Agent. See [Workflows overview](/ai-agent/workflows/overview#where-workflows-live).
## Settings shared by both agent types
Both agent types share the same core configuration. A Workflow Agent simply adds the **Chat Workflows** tab on top. These settings apply regardless of the type you choose:
* **Profile** — agent name, avatar, command, and where the agent responds.
* **Knowledge Base** — content used for RAG answers (and by the AI response node and "Answer with Knowledge base" end action in workflows).
* **Instructions** — the prompts for AI answering. The OpenAI prompt drives free-form answers (including the "Answer with AI" end action and fallback answering); the Knowledge base prompt drives knowledge-base (RAG) answers. (The OpenAI Chat workflow node uses its own per-node instructions instead.)
* **Widget** — appearance and behaviour of the embedded chat widget.
* **Human Handover** — configuration for the Live Chat node and the handover experience.
Keep your instructions and knowledge base up to date for both agent types. For Knowledge Agents they drive every answer; for Workflow Agents they power fallback AI answering and the AI response node.
## Next steps
For Knowledge Agents:
* [Knowledge base](/ai-agent/knowledge-base) — add files, websites, and pages the agent answers from
* [Intake forms](/ai-agent/intake-forms) — collect information before the conversation starts
* [How to use](/ai-agent/how-to-use) — interact with your AI Agent in private and group chats
For Workflow Agents:
* [Workflows overview](/ai-agent/workflows/overview) — understand what a workflow is and when to use one
* [Building a workflow](/ai-agent/workflows/editor) — create, publish, and activate a workflow
* [Workflow nodes](/ai-agent/workflows/nodes) — reference of every available node type
* [Variables, branching & testing](/ai-agent/workflows/variables-and-testing) — personalise flows and test before going live