The most used nodes

Steer how a run proceeds, such as repeating steps over a list, branching, or stopping the run early.

Expose the flow's actions through a server that AI clients can connect to and call as tools.

Call another flow and use what it returns, or let this flow be called by others as a reusable piece.

Change text along the way: find and replace, split, join, trim, and switch upper or lower case.

Pause the run until a person approves or rejects, then continue down the path that matches their answer.

Create, read, convert, zip, and unzip files, and check their details, while passing them between steps.

Reshape data by renaming fields, picking out parts, and building new objects for the steps that follow.

Create a task for a team member and wait for them to review, approve, or do the manual work needed.

Reach many different AI models through a single service to produce text, code, or images.

Work with PDF documents, such as pulling out text or combining, splitting, and creating pages.

Send prompts to a hosted large language model and use the returned text or answers in later steps.

Do basic arithmetic on two numbers (add, subtract, multiply, divide) or produce a random number for later steps.

Send WhatsApp messages, read and update contacts, and drive chat replies through a WAHA server.

Manage email marketing contacts and lists and send campaign or transactional messages to them.

Look up stored account connections at run time so a step can pick credentials based on the data passing through.

Create random passwords and compute hash values from text, for example to sign or compare data.

Turn CSV text into rows and fields a flow can work with, and build CSV text back out of records.

Watch a feed address for newly published items and read their titles, links, and content.

Create or analyse video clips with an AI model and pass the resulting file or description onward.

Start a flow when someone books or cancels a meeting, and read the booking and invitee details.

Generate text and reason over long inputs using Claude models.

Search the web and pull the raw text out of result pages, so a model can read fresh sources.

Read, filter, and edit JSON with path expressions so a flow can pull out or reshape specific fields.

Call models deployed in your own Microsoft cloud subscription for chat, embeddings, and related tasks.

Connect to a self-hosted inference server to run open models for generation, embeddings, and reranking.

Route requests to many commercial and open-source models through one gateway and a single key.

Generate text, create embeddings, and turn text into spoken audio with one hosted model family.

Convert documents such as PDFs into machine-readable markdown or JSON so later steps can extract their content.

Fetch web pages as clean text, search the web, and prepare content for models to read.

Crawl a whole site or scrape single pages and return the page content as structured text.

Run a web search query and get back result snippets along with the matching page content.

Use GLM-family hosted models for chat, text generation, and embeddings.

Pull matching pieces of text out of a string using a pattern.

Turn markdown into DOCX, PPTX, XLSX, PDF, PNG, SVG, HTML, notebook, CSV, JSON, XML or wiki files, and save code blocks as source files.

Read and act on repositories, issues, and other data held in a hosted source code service.

Add embedding and reranking models for building search and retrieval steps.

Post messages into a workplace chat group through its bot.

Send messages to a team group chat through its robot webhook.

Run text embedding models on a fast self-hosted endpoint for search and retrieval steps.

Search the web through a privacy-focused engine and return the results to the flow.

Upload, download, copy, move, share and search files and folders in a Google cloud drive.

Send a prompt, with the incoming data filled in, to a connected language model and return its reply in one straight pass with no tools.

Supply a Google Gemini chat model, with its key and settings, as the brain that an agent, chain or extractor step uses.

Keep the last few turns of a conversation in memory and feed them back to the model so it remembers recent context per chat session.

Send a plain text or HTML message with attachments through any SMTP mail server, without being tied to one mail provider.

Supply a chat model served through OpenRouter, giving steps access to many vendors' models behind one key.

Create, read, update, delete and search records in a hosted database table, so a base can act as input or storage.

Supply an Anthropic Claude chat model, with its key and settings, as the brain that an agent, chain or extractor step uses.

Turn incoming files or text into document chunks that embedding and vector store steps can index or search.

Use Google Gemini directly for single jobs such as text replies, image and video handling, audio transcription and file uploads.

Create and update pages, blocks and database entries in a workspace wiki, and search or read existing content.

Turn text into OpenAI numeric vectors that a vector store or retriever step uses for similarity search.

Pull values out of HTML with CSS selectors, or build and tidy HTML markup from the data flowing through.

Run SQL queries against a Postgres database, and insert, update, upsert, select or delete rows in its tables.

Pull named fields out of messy text with a language model and hand them on as clean structured values.

Create, update, delete and list calendar events, and check when someone is free or busy.

Expose another saved automation as a tool an agent can call with arguments, getting its result back as part of the reasoning.

Send text, template and media messages to customers over the business messaging cloud API and handle incoming ones.

Supply a chat model hosted on Groq, giving steps a fast low-latency brain for agents, chains and extractors.

Create documents and read or insert and replace their text, so written content can be generated and updated automatically.