> For the complete documentation index, see [llms.txt](https://docs.borndigital.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.borndigital.ai/digital-agent/conversation-flow/nodes-explained/ai-node.md).

# AI node

The AI node lets a large language model answer the user, use tools and decide where the conversation goes next.

The AI node sends the conversation to a large language model (LLM) together with your instructions. The model can answer the user, use tools — the knowledge base, web search, MCP servers of other systems — and move the conversation to another node. The answer is stored in a variable, so the node can also work in the background: classify a request, extract data or write a summary.

This section describes **AI node 2.0**, the version every new AI node starts with.

## AI node 2.0 and the old AI node

The editor header has a **2.0** switch.

|                     | AI node 2.0 (switch on)                                                                   | Old AI node (switch off)                           |
| ------------------- | ----------------------------------------------------------------------------------------- | -------------------------------------------------- |
| Used by             | Every new AI node                                                                         | AI nodes created before 2.0, until you switch them |
| Tools               | Knowledge base, Web search, Redirect to human, Transfer to node, End process, MCP servers | One knowledge-base tool                            |
| Files in user input | Yes                                                                                       | No                                                 |
| Reasoning settings  | Yes, for reasoning models                                                                 | No                                                 |

See [Migrating from the old AI node](/digital-agent/conversation-flow/nodes-explained/ai-node/migrating-from-the-old-ai-node.md) before switching an existing node.

## How one turn works

<figure><img src="https://4261467870-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F6iQTvxgRZRwPS1NgIGEb%2Fuploads%2FRqCzfgVLMhUT8a8jNy5W%2Fai-node-diagram-turn.png?alt=media&amp;token=51a25e96-334f-4105-b464-c49275e2d33c" alt="One turn of the AI node"><figcaption><p>One turn of the AI node</p></figcaption></figure>

1. The node builds the input for the model: the **Instructions**, the conversation history (if enabled) and the messages from **Modifying input for AI response** — by default `{current_utterance}`, what the user just said.
2. The model answers, or asks to use one or more tools.
3. The node runs the tools — searches the knowledge base, calls an MCP server — and gives the results back to the model. This repeats until the model answers, at most 10 rounds of tool calls per turn.
4. The answer is stored in the response variable (default `gpt_response`), sent to the user if **Send response to user** is on, and added to the history if **Save in history** is on.
5. The flow continues to the node set in **Next**, or to the node chosen by a transfer tool.

## The editor

<figure><img src="https://4261467870-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F6iQTvxgRZRwPS1NgIGEb%2Fuploads%2FXN4DrguQUc5bUACzJVs2%2Fai-node-essential.png?alt=media&amp;token=4bf89692-e089-40b3-a958-21452b197d3d" alt="The AI node editor, Essential tab"><figcaption><p>The AI node editor, Essential tab</p></figcaption></figure>

* **Header** – the node name (always prefixed `AI_`), arrows to the previous and next node, and the **2.0** switch.
* **Essential** tab – **Behaviour** (the Instructions), **MCP servers** and **Tools**: the settings you change in almost every node.
* **Configuration** tab – input messages, provider and model, temperature and reasoning, history, response handling, streaming and timeout.
* **Footer** – **Next**: the node the flow continues to.

## Quick start

1. Add an AI node from the **Nodes** library (chip **AI**) or with the **+** menu of another node.
2. Name the node, for example `AI_ASSISTANT`.
3. In **Behaviour → Instructions**, describe the role, the scope and the style of the answers.
4. Add the tools the node needs — for example the knowledge base.
5. In the footer, set **Next**. For an assistant that keeps the conversation going, lead it to an ANSWER node that waits for the next user input and leads back to the AI node.
6. Train the project and test the conversation.

## In this section

* [Essential tab](/digital-agent/conversation-flow/nodes-explained/ai-node/essential-tab.md) – Instructions, prompt library, tools overview
* [Tools](/digital-agent/conversation-flow/nodes-explained/ai-node/tools.md) – Knowledge base, Web search, Redirect to human, Transfer to node, End process
* [MCP servers](/digital-agent/conversation-flow/nodes-explained/ai-node/mcp-servers.md) – connecting tools of other systems
* [Indexer MCP server](/digital-agent/conversation-flow/nodes-explained/ai-node/indexer-mcp-server.md) – knowledge-base search over several indexes, hybrid search, labels, pinned indexes, and the variables that control it
* [Configuration tab](/digital-agent/conversation-flow/nodes-explained/ai-node/configuration-tab.md) – input and attachments, provider and model, reasoning, history, streaming, timeout
* [Response variable and routing](/digital-agent/conversation-flow/nodes-explained/ai-node/response-variable-and-routing.md) – what the node stores, where the flow continues, errors
* [Migrating from the old AI node](/digital-agent/conversation-flow/nodes-explained/ai-node/migrating-from-the-old-ai-node.md)

{% hint style="info" %}
See the [prompting cookbook](/for-advanced-users/prompting-cookbook.md) for tips on writing Instructions.
{% endhint %}