---
title: "Flow Agents: AI Agent Node, Knowledge Base, Web Search, API Call and Actions Tools"
description: "Configure the Flow Agents AI Agent node and every tool it can use, including Knowledge Base search, Web Search, API Call with Import cURL, and the CRM Actions platform."
url: https://help.aesthetixcrm.com/en/articles/2234-flow-agents-ai-agent-node-knowledge-base-web-search-api-call-and-actions-tools
collection: "Agent Studio"
updated: 2026-11-20
language: en
---

# Flow Agents: AI Agent Node, Knowledge Base, Web Search, API Call and Actions Tools

Configure the Flow Agents AI Agent node and every tool it can use, including Knowledge Base search, Web Search, API Call with Import cURL, and the CRM Actions platform.

The AI Agent node is the "brain" of a Flow Agent in Agent Studio. It reads the input, follows your prompt, picks tools, captures information, and hands results to the next node. This guide covers every setting on the node and every tool you can attach to it. You build Flow Agents in Aesthetix CRM under **AI Agents > Agent Studio > Flow Agents**.

> **HIPAA caution:** AI features in Aesthetix CRM are not approved for processing Protected Health Information (PHI). Do not put PHI into agent prompts, tool instructions, Knowledge Bases, API request bodies, runtime variables, or test conversations. Do not send PHI to outside systems through API Call or MCP tools. Use general practice information and made-up test data.

Agent Studio is mainly for maintaining existing Flow Agents. New agents are best created as Managed Agents. Everything below applies to the flow-based agents you already have.

## AI Agent Node

### What it does

The AI Agent node runs when it receives input from a Start Trigger or from another connected node. Unlike fixed logic, it adapts to the prompt, the tools available, the user's input, and the context. It decides the next best step: generate a response, use a tool, capture information, or pass output onward. It has five core elements: Prompt, Model, Mode, Tools, and Variables, plus its connections to other nodes.

### When to use it

Use the AI Agent node when the next step depends on context or intent: conversational replies, lead qualification, support triage, Knowledge Base lookup, web search, data extraction, CRM updates, tool selection, and dynamic routing. Use a Sequential node instead when steps must run in a fixed order.

Medical-aesthetics examples: qualify a new inquiry (which treatment, timeline), answer questions about services and hours from your Knowledge Base, and tag the contact for the front desk to follow up. Keep each AI Agent node focused on one primary role; many unrelated jobs on one node are harder to control and test.

### Prompt

The prompt defines the agent's role, tone, tasks, and when to use tools. A strong prompt includes:

![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-cmYHRvgNT61kz6byYpnpcRKLKydkxCQJBtsJqw4uGG95VWOd58Lf138HBMqxQHKzvrfwflLTBCW.png?v=1&k=us1&s=api&e=1795348800&sig=yBDnk-iYZTklPOMcUlYo-wt2XrP7rssKHaDaxlmEjg8)

-   The agent's role.
    
-   The goal of the node.
    
-   What the agent should and should not do.
    
-   Tone or style requirements.
    
-   When to use tools.
    
-   What information to capture.
    
-   What output to pass forward.
    

Example: "You are an inquiry qualification assistant for our practice. Ask clarifying questions when needed, identify which treatment category the person is interested in, capture it as a runtime variable, and use the Knowledge Base tool when the person asks about services, pricing, or hours. Never give medical advice."

**Insert variables in the prompt.** Click the `{}` icon in the prompt field to insert dynamic values. At run time the agent replaces each variable with real data. For example, `{{contact.name}}` in "Greet the user by name and help them with their request" produces "Hi Jordan, how can I help you today?" Add only the variables you need so the prompt stays clear.

![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-i7mNsGQw163LBMgTdB9M92N5ywDtQkqUBoHPiBhFEaC2V6OTBk95ZNSwb2Ma4tw2ydqo6n0Prbc.png?v=1&k=us1&s=api&e=1795348800&sig=NgH3NCBmJ7-yQEq5LOh4sNxoJJH7vPsKpcwN5hm8Btw)

<table style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p><strong>Variable source</strong></p></th><th colspan="1" rowspan="1"><p><strong>What it gives you</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Global Config</p></td><td colspan="1" rowspan="1"><p>System-level settings and shared values that apply across the agent.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Input Variables</p></td><td colspan="1" rowspan="1"><p>Data passed from earlier nodes or triggers.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Runtime Variables</p></td><td colspan="1" rowspan="1"><p>Data the agent extracted during the run.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Account</p></td><td colspan="1" rowspan="1"><p>Business-level details such as company information.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Custom Values</p></td><td colspan="1" rowspan="1"><p>Static values you defined for repeated use.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Right Now</p></td><td colspan="1" rowspan="1"><p>Current date and time, for time-sensitive replies.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Contact</p></td><td colspan="1" rowspan="1"><p>Contact details such as name, email, and phone.</p></td></tr></tbody></table>

**Enhance Prompt.** This option rewrites your prompt into a clearer, more structured version. Review the suggestion, then accept it or reject it to keep your original. You stay in control of the final instructions.

### Model

Choose a model from the **Model** dropdown. Higher-capability models give stronger reasoning for complex tasks and ambiguous intent. Smaller, lighter models are faster and cheaper for simple jobs, and some options favor low latency. A stronger model may recognize "I'm exploring options" as a potential lead, while a lightweight model may answer more generically. Pick the lightest model that meets your quality bar, and test.

![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-DUAKDFkG6k7OIixL7nMyAUhyoXXzJftmcXwg8vYcVX06UcNXf40PcR2QwXv9Q3MleQb7hYWoghq.png?v=1&k=us1&s=api&e=1795348800&sig=rSzVsAb9FoNY8RXm5vZE3vRHFEu5VlOgAfoNuBCfEPQ)

### Mode

<table style="min-width: 75px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p><strong>Mode</strong></p></th><th colspan="1" rowspan="1"><p><strong>Behavior</strong></p></th><th colspan="1" rowspan="1"><p><strong>Example</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Conversational Mode</p></td><td colspan="1" rowspan="1"><p>The agent responds to the user and handles interaction.</p></td><td colspan="1" rowspan="1"><p>Answer a question about treatment hours.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Task Based Mode</p></td><td colspan="1" rowspan="1"><p>The agent performs work silently, with no user-facing reply.</p></td><td colspan="1" rowspan="1"><p>Extract an email address and update the contact record.</p></td></tr></tbody></table>

### Tools

Tools let the agent do more than generate text. Based on the prompt and context, the agent decides when a tool is needed and chooses one. Attach only the tools you need. Add a tool from the **Add a Tool** dropdown inside the node, or drag it from the **Add node** panel into the node.

### Runtime variables

Runtime variables capture structured data during the run, such as a name, email, or request type. The agent uses each variable's Name and Description to know what to look for and extracts it automatically.

![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-CnNV2zfapIX7TUpbEx7dv7VZpyruDW4A0B6xB248QqAWEfK0PVDlPFsfXd0wBu27SwXEz8HGD9J.png?v=1&k=us1&s=api&e=1795348800&sig=hUEfmPeCB-ruPCmhkET8gIzpcUuUBrSqAzHHZREjCTs)

<table style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p><strong>Field</strong></p></th><th colspan="1" rowspan="1"><p><strong>Meaning</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Name</p></td><td colspan="1" rowspan="1"><p>Identifier for the data (for example <code>user_name</code>, <code>email</code>, <code>requirement</code>).</p></td></tr><tr><td colspan="1" rowspan="1"><p>Type</p></td><td colspan="1" rowspan="1"><p>String (text), Number (numeric such as quantity), Boolean (true or false), or JSON (structured data for complex capture).</p></td></tr><tr><td colspan="1" rowspan="1"><p>Description</p></td><td colspan="1" rowspan="1"><p>A clear instruction describing what to extract.</p></td></tr></tbody></table>

For example, "Hi, I'm Jordan and I'd like a consult" can set Name = Jordan and Requirement = Consult for use in later steps. Do not capture PHI into runtime variables. If a variable comes back empty, the user may not have given the value, the description may be unclear, or the prompt may not tell the agent to capture it. Improve both and test again.

### Node connections

The node outputs its result (a response, an action result, or extracted data) to the next connected node, for example sending qualified details to an action that updates the CRM. Make sure the node is connected to the next step.

### Setup steps

1.  Open the agent in **AI Agents > Agent Studio** and open the **Nodes** panel.
    
2.  Add the **AI Agent** node and connect it to a Start Trigger or previous node.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-5r5ZTbhstDxIrRyfEz6KoiayAdJw22BIY7a5izautAAdLo3iIvylhfwjyKTtApTiRB3RcgNfRVV.png?v=1&k=us1&s=api&e=1795348800&sig=5Z65eUwY7PUQnaqLjGx1qBDYr9LIRGAg1NWRr0dm9f0)
3.  Enter a clear prompt. Use the `{}` picker for variables. Optionally click **Enhance Prompt**.
    
4.  Select the **Model**.
    
5.  Choose **Conversational Mode** or **Task Based Mode**.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-Ciy9YDC9TGshm1O0e64D1N9TUh2khiN3rNNExHylgusPMwXxIPmdYGGW5OnMkbO9UYlCk62n6O8.png?v=1&k=us1&s=api&e=1795348800&sig=ZW11prICxriGr5cMZPppDE4XVs6tVANGj85WtZm3fw8)
6.  Attach only the tools needed.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-tSLWk0GKlQxp0jW2b8hX8dRVO0n8K4bSM8klnyRTrvE4Et0RPvgaYf9r8U2rZEOAr2LN3szMXyc.png?v=1&k=us1&s=api&e=1795348800&sig=uYNw3yBfNSqP35fc-Aavq8Pt5EcLwZLho7T8DnacGw0)
7.  Define runtime variables to capture data.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-4midnvLmX6o1Kns3bZbQR6fZYLIcKpwqlNF4ygb36d7EPjBorpqARD1fxh6DNyIaSPrQAG308lR.png?v=1&k=us1&s=api&e=1795348800&sig=2tJsyuwTlJe8b2QDD1RI0ZExlfkKO0btUIxRjwOD8ZE)
8.  Connect the node to the next step.
    
9.  Click **Test**, and test again after any change to the prompt, model, mode, tools, variables, or connections.
    

## Tool Catalog

<table style="min-width: 75px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p><strong>Tool</strong></p></th><th colspan="1" rowspan="1"><p><strong>What it does</strong></p></th><th colspan="1" rowspan="1"><p><strong>When to use it</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Router</p></td><td colspan="1" rowspan="1"><p>Directs the flow to a path based on user intent or agent decisions.</p></td><td colspan="1" rowspan="1"><p>Branching to sales, support, or fallback paths.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Search Knowledge Base</p></td><td colspan="1" rowspan="1"><p>Retrieves answers from selected Knowledge Bases.</p></td><td colspan="1" rowspan="1"><p>FAQs, service details, pricing, policies, troubleshooting.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Search the Web</p></td><td colspan="1" rowspan="1"><p>Fetches real-time information from the internet.</p></td><td colspan="1" rowspan="1"><p>Up-to-date or external questions not in your Knowledge Base.</p></td></tr><tr><td colspan="1" rowspan="1"><p>MCP Server</p></td><td colspan="1" rowspan="1"><p>Connects to external systems through the Model Context Protocol.</p></td><td colspan="1" rowspan="1"><p>Custom tools, databases, or third-party platforms.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Actions</p></td><td colspan="1" rowspan="1"><p>Runs predefined CRM operations.</p></td><td colspan="1" rowspan="1"><p>Update records, send messages, tag contacts, notify the team.</p></td></tr><tr><td colspan="1" rowspan="1"><p>API Call</p></td><td colspan="1" rowspan="1"><p>Sends requests to external systems and reads responses.</p></td><td colspan="1" rowspan="1"><p>Real-time data exchange with other applications.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Text Generation</p></td><td colspan="1" rowspan="1"><p>Creates written content from context.</p></td><td colspan="1" rowspan="1"><p>Messages, summaries, structured outputs.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Image Generation</p></td><td colspan="1" rowspan="1"><p>Creates images from a prompt.</p></td><td colspan="1" rowspan="1"><p>Dynamic visuals.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Video Generation</p></td><td colspan="1" rowspan="1"><p>Creates videos from a prompt, with resolution and aspect ratio settings.</p></td><td colspan="1" rowspan="1"><p>Short video content.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Audio Generation</p></td><td colspan="1" rowspan="1"><p>Converts text to speech with text-to-speech models.</p></td><td colspan="1" rowspan="1"><p>Spoken replies.</p></td></tr></tbody></table>

For API Call, configure the endpoint and use variables to pass dynamic data, which later steps can use. For generation tools, configure the prompt and use variables to personalize output. See the Flow Agents Generative AI Tools and Brand Voice guide for every generation setting, and the Flow Agents Sequential Node, Router, Capture Tools and Always Exit guide for routing.

## Search Knowledge Base Tool

The Search Knowledge Base tool lets the agent answer from the information in one or more Knowledge Bases. A Knowledge Base can hold website pages, FAQs, tables, rich text, uploaded files, and web search content. The tool lives inside an AI Agent node. It does not start the agent; a trigger does.

### Knowledge Base overview

Knowledge Bases are created and managed at **AI Agents > Knowledge Base**. Each one opens to a **Summary** view with an overview of its connected sources and structure. Data sources appear as tabs, each showing the number of connected items and actions to add or manage content:

![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-3TqxUIoacgMZE08WxuMn5W6MSVL03ubtagQvVu1aVxB14yAxEPjmrQvDNZ2koTJk1RpP2kCVXkv.png?v=1&k=us1&s=api&e=1795348800&sig=S_G4PSkbjFBBACRosDCxCqhMyNbexUF-FxfpkmrHIDQ)

<table style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p><strong>Source tab</strong></p></th><th colspan="1" rowspan="1"><p><strong>Use it for</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Web Crawler</p></td><td colspan="1" rowspan="1"><p>Crawling website pages so the Knowledge Base can use website content.</p></td></tr><tr><td colspan="1" rowspan="1"><p>FAQ</p></td><td colspan="1" rowspan="1"><p>Direct question-and-answer content.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Web Search</p></td><td colspan="1" rowspan="1"><p>Adding web-based information sources.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Tables</p></td><td colspan="1" rowspan="1"><p>Structured CSV data such as pricing, product lists, or service details.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Rich Text</p></td><td colspan="1" rowspan="1"><p>Manually written content such as documentation, policies, or instructions.</p></td></tr><tr><td colspan="1" rowspan="1"><p>File Upload</p></td><td colspan="1" rowspan="1"><p>Supported document files.</p></td></tr></tbody></table>

You can add multiple sources under one tab where supported. Admins and editors can manage Knowledge Bases; viewer-only users see no management controls. The refreshed layout is rolling out gradually, so you may not see the Summary view or tabs on every account yet; they appear automatically. The refresh does not change how content is written, permissions, indexing, AI behavior, or existing Knowledge Bases. Creating a Knowledge Base does not connect it to an agent; you must attach it. Use the Knowledge Base Retrieval Tester to ask sample questions and see which sources come back.

Only add practice-level information (services, pricing, hours, policies, aftercare in general terms). Do not upload documents that contain patient records or PHI.

### Set it up

1.  Create or select a Knowledge Base at **AI Agents > Knowledge Base** and add source content. Confirm it covers the questions you expect.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-xfHuVnLWSn8tH4Sn9v89qIa2S6ZmKNOhYZDN8382oTTyEK9TrQIhMFTO8Ci9hQ7NIOYu9hcKekK.png?v=1&k=us1&s=api&e=1795348800&sig=89EqIYcGxjF_xVnHKV4zcGWsUvc50KJLFw2dohUOKxM)
2.  Go to **AI Agents > Agent Studio**, create or open an agent, and open the canvas.
    
3.  Add a trigger to start the agent. Common choices: **Chat Message** (the most common for this tool), **Form Submitted** (the agent can review the submission and use the Knowledge Base to suggest next steps), and **Lead Tag** (adding or removing a tag starts the agent, which can supply tag-related guidance).
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-U4yD30AVjsHK3Ycg3O2aBLWrBIIFVePb9lSufdYfnQU1OPDAlibJbjD87YgIb0hBPtBry5suaKs.png?v=1&k=us1&s=api&e=1795348800&sig=HuGnCEqcFlFrF8fVqgxhdopLLU5fhups_9NpVzxl4mY)
4.  Add an **AI Agent** node and open its settings.
    
5.  Add **Search Knowledge Base** from the **Add a Tool** dropdown, or drag it from the **Add node** panel into the node.
    
6.  Configure it:
    
    -   **Instructions** (required): tell the agent when and how to use the tool.
        
    -   **Enhance Instructions:** improves or expands the instruction before saving.
        
    -   **Variable picker:** inserts dynamic values into the instruction.
        
    -   **Knowledge Base Selection:** select one or more Knowledge Bases to search. Choose the ones with the most relevant information.
        
    -   **Cancel** closes without saving. **Save** stores the selection and instructions.
        

### Test it

1.  Open the agent and run a test conversation.
    
2.  Ask a question the Knowledge Base should answer and confirm the answer matches the source.
    
3.  Ask an unrelated question to confirm the agent does not overuse the Knowledge Base.
    
4.  Adjust the tool instruction if it is used too often or not enough, then save and test again.
    

If the answer is not in the Knowledge Base, tell the agent in its instructions what to do (for example, offer to connect the person with your team).

## Web Search Tool

The Web Search tool lets the agent fetch real-time information from the internet when your Knowledge Base does not cover a question. The agent turns the user's question into a search query, retrieves results, and builds an answer from them.

### Settings

<table style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p><strong>Element</strong></p></th><th colspan="1" rowspan="1"><p><strong>What it does</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Instructions</p></td><td colspan="1" rowspan="1"><p>Defines how and when to use the tool, for example only when the Knowledge Base lacks the answer or when real-time data is needed. This limits unnecessary searches.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Query</p></td><td colspan="1" rowspan="1"><p>What the agent searches for. Prefer a dynamic query built from a runtime variable instead of fixed text, so each search reflects the user's question. A fixed query returns the same results every time.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Variables</p></td><td colspan="1" rowspan="1"><p>Insert runtime variables with the <code>{}</code> picker so the query adapts to user input.</p></td></tr></tbody></table>

### Configure it

1.  Open the AI Agent node, click **Add a Tool**, and select **Search the Web**.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-D8ACEEKHX9AzCtY3aol3xHiqmKrbkATDKeh0Q4oVQiP7bttH1jilnjNLoaxLJA7lZkbhMIilJrA.png?v=1&k=us1&s=api&e=1795348800&sig=6aohQm_ktEU7ShdNaqZctn5jxsSMFx2ivppSOIZc9yE)
2.  Add **Instructions**, for example "Search the web only when the Knowledge Base does not contain the answer."
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-gis2vhxWtC03kmFKRTsF9m0kkcYu9AqI1SP6M8C7yXiIUkmc3aD7QL5dc3YW4QsNSQgxIlCJxva.png?v=1&k=us1&s=api&e=1795348800&sig=4y8gYsj-w-Yf0b05OZvlaZdxTzH-S9xzDbirUJubNBM)
3.  Set the **Query**, using the `{}` picker to insert the runtime variable holding the user's question.
    
4.  Test with different inputs and check that results are relevant and accurate.
    

### Guidance and limits

-   Use the Knowledge Base for controlled, consistent information about your practice. Use Web Search for information not held internally or that changes.
    
-   Web results depend on outside sources, so answers can vary between runs.
    
-   You can guide the search with instructions and the query, including restricting to certain topics or types of sources.
    
-   Never put patient details into a web search query. Search queries leave your account.
    
-   Do not rely on web search for clinical or medical claims about treatments; keep medical claims in approved Knowledge Base content reviewed by your team.
    

## API Call Tool and Import cURL

The API Call tool lets an agent make an external web request while it runs, to retrieve data, send information, or trigger an outside process. You can add it as a tool inside an AI Agent node, or as a processing step in a Sequential node.

### API Call tool vs public APIs

<table style="min-width: 75px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p></p></th><th colspan="1" rowspan="1"><p><strong>API Call tool</strong></p></th><th colspan="1" rowspan="1"><p><strong>Agent Studio public APIs</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Direction</p></td><td colspan="1" rowspan="1"><p>The agent calls an external API during its flow.</p></td><td colspan="1" rowspan="1"><p>An external application lists, retrieves, and runs production-ready agents without logging in.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Use when</p></td><td colspan="1" rowspan="1"><p>The agent needs outside data or must trigger an outside system.</p></td><td colspan="1" rowspan="1"><p>An outside system needs to interact with your agents.</p></td></tr></tbody></table>

### Import cURL

Import cURL turns a pasted cURL command into a configured API request so you do not copy each value by hand.

**What it fills in:** HTTP method, endpoint URL, headers, and request body. It detects body types JSON, HTML, XML, Text, and Form Data. It understands `--location` redirects, `-u` basic authentication, multiline request bodies, and backslash line continuations.

**Steps:**

1.  Go to **AI Agents > Agent Studio**, then create or open an agent.
    
2.  Add or select a start trigger such as **Chat message**, **Form submitted**, **Lead tag**, or **Workflow**.
    
3.  Add an **AI Agent** node or **Sequential** node.
    
4.  Add the **API Call** tool or processing node and open its configuration panel.
    
5.  Click **Import cURL**, paste the command into the **cURL Command** field, and click **Import**.
    
6.  Review the populated fields: HTTP method, endpoint URL, headers, request body, body type, variable name, and Instructions (when inside an AI Agent node).
    
7.  Edit as needed and click **Save**.
    
8.  Click **Test**, then **Publish** when ready.
    

**Review before you save:**

<table style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p><strong>Field</strong></p></th><th colspan="1" rowspan="1"><p><strong>Check</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Instructions</p></td><td colspan="1" rowspan="1"><p>Describes what the call does and when the agent should use it.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Variable Name</p></td><td colspan="1" rowspan="1"><p>Where the result is stored.</p></td></tr><tr><td colspan="1" rowspan="1"><p>API Endpoint</p></td><td colspan="1" rowspan="1"><p>The method and URL are correct.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Headers</p></td><td colspan="1" rowspan="1"><p>Required headers such as content type or authorization are present.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Request Body</p></td><td colspan="1" rowspan="1"><p>Payload and format are right.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Authentication values</p></td><td colspan="1" rowspan="1"><p>Tokens, credentials, or keys are replaced, removed, or secured. Do not leave real secrets in pasted examples.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Variables</p></td><td colspan="1" rowspan="1"><p>Add Agent Studio variables where the request should use contact, conversation, or runtime data.</p></td></tr></tbody></table>

### Two-way sync

When you reopen the Import cURL window after configuring a call, the current configuration is shown rebuilt as a cURL command. Use it to verify the format, compare against the outside API's documentation, edit headers or body quickly, and troubleshoot, then re-import the updated command.

### Troubleshooting a request

Check the endpoint URL, HTTP method, headers, authorization values, request body, body type, and any variables. Then test again. Always test before publishing, especially with authentication, variables, headers, or a body. Do not send PHI to an external API.

## Actions Platform

The Actions Platform is a set of built-in CRM actions that Flow Agents can run in real time, so an agent can take trackable steps (send messages, tag contacts, update fields) instead of only talking. No custom code or outside integration is needed. Actions are available inside action-capable nodes such as AI Agent and Sequential nodes, where you pick the CRM operation and map variables from the conversation.

### Out-of-the-box actions

<table style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p><strong>Action</strong></p></th><th colspan="1" rowspan="1"><p><strong>What it does</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Send SMS</p></td><td colspan="1" rowspan="1"><p>Texts the current contact through your configured phone provider. Good for confirmations, reminders, and follow-ups. SMS usage follows your existing phone pricing and A2P compliance rules.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Send emails (with templates)</p></td><td colspan="1" rowspan="1"><p>Sends transactional or nurture email, optionally from an existing email template, with subject and body snippets personalized through variables.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Tag a contact</p></td><td colspan="1" rowspan="1"><p>Applies tags based on conversation outcomes, such as Hot Lead or Needs Follow-Up, so workflows, pipelines, and reports can react.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Create new tags</p></td><td colspan="1" rowspan="1"><p>Lets the agent create tags that do not exist yet, if you enable that behavior.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Create custom fields and custom values</p></td><td colspan="1" rowspan="1"><p>Stores structured data from AI conversations (for example Preferred Treatment or Budget Range) as persistent CRM data.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Send internal notifications</p></td><td colspan="1" rowspan="1"><p>Alerts team members when something important happens, such as a lead asking to talk to the front desk.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Generate content</p></td><td colspan="1" rowspan="1"><p>Drafts a blog post, social caption, or ad copy, then stores it in a field or note, sends it for review, or emails it.</p></td></tr></tbody></table>

Example SMS: "Thanks for booking a consult, {first\_name}! Here is your confirmation link: {calendar\_link}". Messages the agent sends must follow your consent and opt-in rules, and must not include PHI.

### Agentic and sequential use

An AI node interprets intent, a Router chooses the branch, and an action-capable node then runs the exact CRM action. Examples:

-   If a lead shows high intent, tag them and send an instant SMS with a booking link.
    
-   If a patient asks for a refund, notify the team and apply a "Refund Requested" tag for reporting.
    
-   If someone asks for pricing, generate a personalized follow-up email that recaps their question.
    

Sequential nodes group steps that must run in order, such as set custom fields, tag the contact, then send an email. They can mix reasoning steps (AI routing) with concrete actions, then return to the main conversation flow. Examples:

-   **New lead sequence:** create or update the contact, apply a New Lead tag, send an SMS confirmation, and notify the team if interest level crosses a threshold.
    
-   **Post-visit follow-up:** ask how satisfied the person was; for low satisfaction, route to a human and apply an escalation tag; for high satisfaction, send a review request email and add a Review Nurture tag.
    

### Content generation patterns

-   **Generate and send:** draft content and email it for approval.
    
-   **Generate and store:** save long-form copy in custom fields or notes for reuse, or save variations tagged by campaign for testing.
    
-   **Generate and branch:** use AI routing to choose which content type to create, then run the matching action.
    

Humans stay in control of publishing and final approval.

### Triggers and real-time starts

Actions run whenever an agent runs, so triggers decide when the agent acts. Agent Studio triggers include Form Submitted, Lead Tag Added or Removed, and Chat Message. You can also:

-   Invoke agents from workflows (**Marketing > Workflows**) with the **Invoke Agent Studio Agent** workflow action, so a workflow can hand complex reasoning to an agent.
    
-   Route Ask AI queries into agents so chats can start multi-step processes.
    

Any trigger that starts an agent can start Action operations.

### Set up actions

1.  Confirm you have access to **AI Agents > Agent Studio** and that your phone and email setup already works so outbound actions can succeed.
    
2.  Open or create an agent (optionally from a template).
    
3.  Decide where on the canvas the agent should act, such as the end of a Sequential node, after a Router picks an intent, or after collecting key data.
    
4.  Add the action-capable node and choose the action type (Send SMS, Send Email, Tag Contact, Create Custom Field, Send Internal Notification).
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-iGVbM0wD5sd8kLY4ePslmpVvPGbHIort1xTz3NTzScN3el2gej5E1zGj0DNG9uA8Ssc7YyHmfeX.png?v=1&k=us1&s=api&e=1795348800&sig=S0-H1tvMw7ENFk8fOBU-qz4dd3Xi42JqtmP-3J9frjE)
5.  Map variables and custom values to fill in fields such as name, email, and phone, for example setting an "Interest Level" field from a captured value.
    
6.  Attach or confirm triggers on the Start node. To let existing workflows call the agent, add the **Invoke Agent Studio Agent** action and point it at the published agent. To use Ask AI, connect the agent through the Ask AI and Agent Studio integration.
    
7.  Use the Test tab and the execution timeline to simulate the trigger and confirm messages, tags, fields, and notifications fire with the right data.
    
8.  Publish, then monitor contact timelines, workflow logs, and Agent Studio execution logs.
    

### Limits and notes

-   Actions run per trigger or per agent execution, which suits one-to-one real-time automation. For bulk updates, imports, or mass tagging, use workflows, lists, or import tools instead.
    
-   You decide which nodes and actions an agent has. Gate actions behind Router logic, conditions, or variables so tags, notifications, and field updates happen only when criteria are met.
    
-   Actions complement workflows; they do not replace them.
    
-   If the agent books appointments, it works only with native Aesthetix CRM calendars, not EMR calendars.
    
-   If you do not see the options to configure actions, your account may not have access to Agent Studio and AI Employee yet.
    

## Frequently Asked Questions

**When should I use the AI Agent node instead of a Sequential node?**

Use the AI Agent node for context-aware decisions, conversation, tool selection, or data extraction. Use a Sequential node when steps should run in a fixed order.

**What is the difference between Conversational and Task Based mode?**

Conversational Mode sends user-facing replies. Task Based Mode does background work, such as extracting data, without replying.

**How does the node decide which tool to use?**

It uses the prompt, the attached tools, the user input, and the context. Clear tool instructions improve selection.

**Does the Search Knowledge Base tool start the agent?**

No. A trigger starts the agent. The tool is used inside the AI Agent node when a relevant question comes up.

**Where do I create a Knowledge Base?**

At **AI Agents > Knowledge Base**. Then attach it to the agent through the Search Knowledge Base tool.

**When should I use Web Search instead of the Knowledge Base?**

Use the Knowledge Base for controlled, consistent practice information. Use Web Search for information you do not hold internally or that changes often.

**Why are Web Search results sometimes inconsistent?**

They depend on outside sources and the query, so results can change over time.

**Does Import cURL support multiline commands and different body types?**

Yes. It handles multiline bodies, backslash continuations, and JSON, HTML, XML, Text, and Form Data bodies.

**Is Import cURL the same as the Agent Studio public APIs?**

No. Import cURL configures an outgoing call from inside an agent. Public APIs let outside applications run your agents.

**Does the Actions Platform replace workflows?**

No. It complements them. Workflows can call agents with the Invoke Agent Studio Agent action when you need AI reasoning plus CRM actions.

**How are SMS and email charges handled for agent actions?**

Messages are billed through your existing phone and email setup. Actions do not change your carrier or email pricing.

**Can I limit which actions an agent can take?**

Yes. You choose the nodes and actions in the flow, and can gate them behind Router logic, conditions, or variables.

**Can I use patient information in prompts, the Knowledge Base, or API calls?**

No. AI features are not approved for PHI. Use general practice information and made-up examples only.
