BusyBot
Search all guides

Reference · Nodes

AI nodes

AI nodes put model calls inside a workflow, one step at a time. OpenAI Chat, Claude Chat, Gemini Chat and Grok Chat send a prompt and return text; the structured-output nodes return JSON that matches a schema you define; and vision, speech, image, embedding and web-search nodes cover the other kinds of model work. Simple LLM adds a plain text-generation step and runs without any provider credential.

64 nodes. Every page documents the shipped node: real parameters, real outputs, real error behaviour.

Model calls as workflow steps

Each provider has its own family of nodes, so a workflow can call the model that suits each step instead of one model for everything.

The chat nodes, OpenAI Chat, Claude Chat, Gemini Chat and Grok Chat, send a message, with optional system instructions, and return the model’s text. When the next node needs fields rather than prose, use a structured-output node instead: OpenAI Structured Output, Claude Structured Output, Gemini Structured Output or Grok Structured Output return JSON that matches the schema you give them. For problems with several steps of reasoning, Claude Thinking, Gemini Thinking, OpenAI Reasoning and Grok Reasoning use each provider’s extended-reasoning models.

The rest of the category covers model work other than text. OpenAI Speech-to-Text transcribes audio, and OpenAI TTS and Gemini TTS turn text into speech. Claude Vision and Gemini Vision describe images. OpenAI Embeddings and Gemini Embeddings turn text into vectors for similarity search. OpenAI Web Search, Claude Web Search, Gemini Google Search and Grok Web Search answer from current web results with their sources. For large jobs, Claude Batch and OpenAI Batch submit many requests as one asynchronous job.

Triggers

There are no triggers in this category. Start an AI workflow with a trigger from another category, with Webhook, or by hand with Manual Trigger.

Example workflows

  • Classify support tickets. Claude Structured Output classifies each ticket into a category with a priority and a summary, and Switch dispatches it to a team based on that priority.
  • Transcribe and summarize. OpenAI Speech-to-Text transcribes an MP3 recording to text, and Gemini Chat summarizes the transcript with a Gemini Flash model.
  • Research into sections. Gemini Deep Research runs a competitive analysis of a market segment and returns a report with sources, and Response Parser splits the report into sections by its markdown headers.

All ai nodes

AI Transform Use OpenAI to transform workflow data via natural language instructions. Supports batch code generation (LLM writes JS executed in a sandbox) and per-item direct transformation modes. Action Claude Batch Submit many Claude message requests as one asynchronous Message Batches job at half the standard price, then check its status or cancel it; each input item becomes one request. Action Claude Chat Send a user message to Anthropic's Claude Messages API and return the reply text, with configurable system instructions and sampling settings, one call per input item. Action Claude Citations Ask Claude about inline text, a binary document such as a PDF, or a previously uploaded Anthropic file, and get an answer with citations that point back to the exact source passages. Action (binary) Claude Code Execution Let Claude write and run Python in Anthropic's sandbox to compute, analyze data or process text, returning both the code it wrote and the execution output. Action (binary) Claude File Upload Upload a binary file to Anthropic's Files API and return the file object plus a file ID that downstream Claude nodes can reference as a document or image. Action (binary) Claude Memory Manage conversation memory — stores and retrieves context across messages using multi-turn conversation patterns. Action Claude Search Results Answer a query with Claude while grounding it in search results you supply; each result is passed as a document block with its title, URL and snippet. Action Claude Structured Output Get a JSON object from Claude that matches a schema you define, by forcing a tool call whose input schema is the desired structure; one call per input item. Action Claude Thinking Use Claude extended thinking for complex reasoning with visible thought process. Action Claude Vision Analyze an image with Claude and return its text findings; the image can come from binary data, a URL or an Anthropic file ID from Claude File Upload. Action (binary) Claude Web Fetch Fetch the content of a URL and send it to Claude with an instruction such as summarizing or extracting, returning the response together with the source URL. Action Claude Web Search Have Claude search the web with Anthropic's web search tool and return a synthesized answer plus the source URLs, titles and snippets it drew on. Action Embedding Similarity Compute the similarity score between two embedding vectors using various metrics like cosine similarity or Euclidean distance. Action Gemini Batch Submit multiple content generation requests to Google Gemini in a single synchronous API call, receiving all responses at once. Action Gemini Chat Send a user message to Google Gemini models, optionally with file URIs, and receive a generated response with configurable settings. Action Gemini Code Execution Write and execute Python code in a sandboxed environment using Gemini to solve problems, perform calculations, and analyze data. Action Gemini Deep Research Conduct deep research using Gemini — submits a research query and returns a comprehensive report with sources. Action Gemini Embeddings Generate dense vector representations of text using Google AI's Gemini embedding models, with configurable task types and optional titles for document tasks. Action Gemini File Search Query and analyze the content of files previously uploaded to Gemini, using their file references. Action Gemini File Upload Upload binary files to Google's file service for reference in Gemini model prompts. Action (binary) Gemini Google Maps Find places, get directions, and geocode locations using Google Maps data through Gemini. Action Gemini Google Search Perform web searches and receive synthesized answers grounded in real-time Google Search results, including source URLs and titles. Action Gemini Image Generation Generate images directly from text descriptions using Google Gemini models, producing visual content based on your prompts. Action (binary) Gemini Music Gen Generate music from a text description with Google's Lyria model through Gemini, returning the composition as an audio file. Action (binary) Gemini Structured Output Force Gemini to return valid JSON matching a provided schema, ensuring structured data extraction from text. Action Gemini Thinking Solve complex problems with Google Gemini's thinking mode, revealing a step-by-step chain of thought for transparent deliberation. Action Gemini TTS Convert text into spoken audio using Gemini models and a selection of distinct voices. Action (binary) Gemini URL Context Enable Gemini to fetch and analyze content from URLs provided in a message, grounding its responses in actual web page information. Action Gemini Video Gen Generate videos from text prompts using Google Veo models, with configurable aspect ratio, duration, and person generation controls. Action (binary) Gemini Vision Send an image and a text prompt to Google Gemini to get a text analysis of the image content. Action (binary) Google Cloud Natural Language Analyze the emotional tone of text documents, including per-sentence scores, using Google Cloud Natural Language. Action Grok Batch Create, check the status of, or cancel asynchronous batch jobs for Grok chat completions, processing multiple requests efficiently. Action Grok Chat Send a message to xAI Grok models and receive a text response, with options for web and X search. Action Grok Code Interpreter Write and execute code in a sandboxed environment using Grok to perform calculations, analyze data, and generate outputs. Action Grok Collections Manage document collections on the xAI platform for retrieval-augmented generation, including creating, listing, retrieving, and deleting them. Action Grok Collections Upload Send binary file data, such as PDFs or text, to a Grok collection to enhance Retrieval-Augmented Generation. Action (binary) Grok Image Edit Edit images using xAI Grok by providing a source image and a text prompt describing the desired modifications, returning the edited image(s). Action (binary) Grok Image Generation Generate images from text prompts using xAI Grok Image Generation, returning either binary data or a URL for the generated image. Action (binary) Grok Reasoning Perform complex multi-step tasks with xAI Grok models that generate internal chain-of-thought reasoning before responding, including optional web and X search. Action Grok Structured Output Obtain JSON responses from xAI Grok that conform to a specified schema, ensuring valid and structured data output. Action Grok Video Gen Generate videos from text descriptions or starting images, with adjustable duration, using xAI Grok. Action (binary) Grok Vision Send an image and text prompt to xAI Grok to receive a text analysis of the image content. Action (binary) Grok Web Search Search the web using xAI Grok models, receiving synthesized answers with URL citations and optional domain filtering. Action Grok X Search Search live X/Twitter posts using xAI Grok's built-in search tool, returning a synthesized response with X post citations. Action OpenAI Audio Translation Translate spoken audio from any supported language into English text, accepting various audio formats and returning the translation in a specified text format. Action (binary) OpenAI Batch Create, check, or cancel asynchronous batch jobs for OpenAI chat completions or embeddings. Action OpenAI Chat Generate text completions from OpenAI language models, supporting initial instructions, temperature, and file attachments. Action OpenAI Code Interpreter Execute code in a sandboxed environment via OpenAI Responses API code interpreter tool. Action (binary) OpenAI Embeddings Generate dense vector representations of text using OpenAI models for tasks like semantic search and similarity comparison. Action OpenAI File Search Search through documents stored in OpenAI vector stores using the Responses API to retrieve relevant file content and generate a response with citations. Action OpenAI File Upload Upload files to OpenAI for use in various AI tasks, including assistants, fine-tuning, batch processing, and vision analysis. Action (binary) OpenAI Image Edit Modify an existing image using OpenAI's DALL-E and GPT-Image models based on a text prompt, optionally with a mask for DALL-E 2. Action (binary) OpenAI Moderation Classify text for harmful content categories using OpenAI's moderation models, returning flags and confidence scores for each input item. Action OpenAI Reasoning Use the GPT-5.x frontier models in extended-reasoning mode for complex multi-step tasks. Action OpenAI Speech-to-Text Transcribe audio files into text using OpenAI's transcription service. It accepts various audio formats and returns transcriptions in specified formats, supporting language hints and timestamp granularities. Action (binary) OpenAI Structured Output Force an OpenAI model to return JSON output that strictly conforms to a provided schema, optionally extracting data from an attached file. Action OpenAI TTS Convert text into natural-sounding audio using OpenAI's Text-to-Speech API, with configurable voices, formats, and speed. Action (binary) OpenAI Vision Analyze images using OpenAI's natively multimodal models, supporting various image sources and configurable output. Action (binary) OpenAI Web Search Search the web for information using OpenAI models, returning synthesized answers with source annotations. Action Prompt Template Build prompts from templates with variable substitution using {{variable}} syntax. Action Response Parser Extract structured data from AI responses — parse JSON, extract code blocks, split sections. Action Simple LLM Send each item to OpenAI's Responses API with your instructions and return the model's text, with configurable model settings and optional built-in web search. Action Structured Output Extract specific data from text into a defined JSON structure using AI, ensuring reliable output for arrays, objects, and typed fields. Action

Frequently asked questions

Can an AI node start a workflow?

No. This category has no triggers: AI nodes run inside a workflow once something has started it. Pair them with a trigger from another category, such as Gmail Trigger or Webhook, start the run by hand with Manual Trigger, or put the workflow on a schedule.

Which credentials do AI nodes use?

Each provider's nodes share one credential type holding an API key from your account with that provider: OpenAI for the OpenAI nodes, AI Transform and Structured Output, Anthropic for the Claude nodes, Google AI for the Gemini nodes, and xAI for the Grok nodes. Create it once and every node from that provider can select it. Simple LLM, Prompt Template, Response Parser and Embedding Similarity need no credential.

How do AI nodes combine with core nodes?

Prompt Template builds the prompt from each item's fields, the model node does the work, and Response Parser pulls JSON or code blocks out of a reply when you did not use a structured-output node. If or Switch can then route on the result, for example sending the items OpenAI Moderation flagged down a separate branch.

Last reviewed . Spotted something wrong? Tell us.