360+ built-in commands

Send a prompt to an LLM and store the reply.

Summarize text with an LLM.

Extract structured fields from text with an LLM and store them as a JSON object (use {{{var[field]}}} to read each one).

Classify text into exactly one of the categories you give, and store the chosen label (route the workflow with an If).

Translate text into a target language with an LLM, keeping the original formatting.

Give the LLM a goal and let it reason step-by-step, calling safe built-in tools (calculator, current time, optional web fetch) until it produces an answer.

Send a prompt to Anthropic's Claude models (native Messages API) and store the reply.

Analyse the sentiment of text and store an object with {sentiment, score, reason}. Read it as {{{var[sentiment]}}}.

Answer a question grounded in your own documents. The command chunks the sources, retrieves the most relevant passages (offline TF-IDF) and asks the LLM to answer using only those passages — reducing hallucination.

A tool-using agent that can search your knowledge base and call your own HTTP API tools (plus a calculator and clock) while reasoning toward an answer. Great for RAG over docs and for integrating external services.

Generate natural speech audio from text with ElevenLabs and save it as an MP3. Requires an ElevenLabs API key.

Generate music with Suno via a Suno API gateway (e.g. the open-source suno-api). Saves the resulting audio file. Set the gateway base URL and, if it needs one, an API key.

Generate a video from a text prompt with Google Veo via the Gemini API, wait for it to finish, and download the MP4. Requires a Google AI Studio (Gemini) API key with Veo access.

Generate text with Google Gemini (generateContent). Useful as a step in Google AI workflows. Requires a Gemini API key.

Generate an image from a text prompt with xAI Grok and save it. Requires an xAI API key.

NotebookLM-style grounded synthesis: feed your sources and a research focus, and the LLM produces a cited briefing (summary, key findings, an outline and follow-up questions) using ONLY those sources. Works with any LLM endpoint.

Chunk your documents and save a reusable knowledge index (.json) for fast offline retrieval. Build once, then Search it many times without re-reading the source files.

Retrieve the most relevant passages from a saved knowledge index (offline TF-IDF). Returns the passages joined as text — feed them to an AI Chat prompt for grounded answers.

Compute an embedding vector for text via an OpenAI-compatible /embeddings endpoint (works with OpenAI or a local Ollama model such as nomic-embed-text). Stores the vector as a list.

Send a fully-configurable HTTP request: query, headers, auth, typed body, retries, typed response or download to file.

Send a JSON request and store the parsed JSON response.

Send an authenticated JSON request with a Bearer token.

Extract a value from a JSON variable or text by path (a.b[0]).

Create a .zip from files, wildcards or whole folders.

Extract a .zip archive into a destination folder.

A single node that checks a condition and has two outputs — TRUE and FALSE. Wire each output to the commands for that branch (no separate Else / End If commands needed).

A single node that switches on a value and has one output per case (listed below) plus a Default output. Wire each output to the commands for that case (no separate Case / End Switch commands).

Convert Markdown text or a .md file to a .docx Word document, with an optional Obsidian-style colour theme.

Render Markdown (text, file, folder or wildcard) to HTML with an optional Obsidian-style theme and code highlighting.

Parse CSV (text or file) into a DataTable.

Serialize a DataTable to CSV text and/or a .csv file.

Parse XML (text or file) into a nested object/JSON variable.

Extract text or an attribute from HTML (text or file) using a CSS selector.

Write a message to the local run log.

Extract part of a string by start index and length.

Split text by a delimiter into a list.

Check whether text contains / starts with / ends with a value.

Count the words in a string.

Get the number of characters in a string.

Extract text before, after, or between marker(s).

Extract the first regex match (or a capture group) from text.

Return all regex matches in text as a list.

Replace regex matches in text with a replacement (supports \1 groups).

Extract a value from a JSON string by path (e.g. a.b[0].c).

Parse a JSON array from text or a .json file into a list variable.

Expand a JSON object's top-level keys into individual variables.

Add or subtract time from a date and format the result.

Format a date or number using a format string.

Extract text from a PDF file (requires the 'pypdf' package).

Create a DataTable with the given columns.

Load a DataTable from CSV text or a .csv file (delimiter auto-detected, BOM-tolerant).

Write a DataTable to a .csv or .json file.

Append a row to a DataTable.

Read a whole row (as a dictionary) into a variable.

Create a new DataTable with rows matching a column condition.

Count the rows in a DataTable.

Read a single cell into a variable.

Set a single cell value.

Remove a row by index.

Sort rows by a column (numeric when possible, else text).

Expand a row's cells into individual variables named by column.

Open a database connection from a SQLAlchemy URL.

Run a query; SELECT results are stored as a DataTable.

Run a non-query statement (INSERT/UPDATE/DELETE/DDL).

Create a new dictionary, optionally pre-filled.

Add or update a key/value pair in a dictionary.

Set the value for an existing or new key.

Read the value for a key into a variable.

Store whether a key exists ('True'/'False').

Find the first key whose value matches.

Store all keys as a list.

Remove a key from a dictionary.

Parse a JSON object string into a dictionary variable.

Serialize a dictionary to a JSON string and/or a .json file.

Store the dictionary's values as a list.

Download a video/audio from 1800+ sites (YouTube, TikTok, Facebook, Instagram, X, plus Chinese sites: Bilibili, Douyin, Youku, Weibo, iQiyi, Xiaohongshu…) via yt-dlp.

Fetch a video's metadata (title, uploader, duration, formats) without downloading it.

Throttle the whole run: wait this many milliseconds before EVERY subsequent command (set once, applies until changed). For a single one-off pause, use Delay instead.

Store a named engine preference value.

Log all current (non-internal) variables and their values.

Start, stop or restart a timer; store elapsed milliseconds.

Choose whether the workflow stops or continues on a command error.

A container with Try, Catch and Finally zones. Commands in Try run first; if one errors, the Catch zone runs ('Error.Message' is set); the Finally zone always runs.

Raise an error, caught by an enclosing Try/Catch block.

Open an existing .xlsx file as a named instance.

Create a new empty in-memory workbook as a named instance.

Save the workbook to a specific file path.

Save the workbook to its current file path.

Read a range as a DataTable using the first row as headers.

Write a list / list-of-lists / DataTable starting at a cell.

Read a cell value (by A1 or row/column) into a variable.

Write a cell value (by A1 or row/column).

Close a workbook instance and discard it from memory.

Store whether a named Excel instance exists.

Get workbook info (path, sheet count, sheet names) as a dict.

Add a new worksheet to the workbook.

Add a new worksheet and make it the active sheet.

Set the active worksheet by name.

Copy a worksheet to a new sheet.

Delete a worksheet by name.

Rename a worksheet.

Store whether a worksheet exists.

Store the name of the active worksheet.

Get a worksheet's dimensions (rows, columns) as a dict.

Store the list of worksheet names.

Clear a cell's value (by A1 or row/column).

Store whether a cell is non-empty.

Record the active cell selection (A1) for the worksheet.

Get the recorded active cell address.

Apply number format / bold / fill to a cell.

Read a cell's number format string.

Read a range into a list of rows (each row a list of cells).

Copy values from one range to a destination top-left cell.

Split a range into a list of per-column lists.

Get the index of the last used row.

Get the index of the last used column.

Get the last non-empty column index within a given row.

Append a row of values after the last used row.

Write values across a specific row starting at column 1.

Append a value to the next empty row in a column.

Delete a row by index.

Delete a column by index (shifts later columns left).

Insert blank row(s) at a position, shifting existing rows down.

Insert blank column(s) at a position, shifting columns right.

Read the used range into a DataTable (first row = headers).

Find the first cell containing a value; store its address.

Filter rows (first row = headers) by a column condition into a DataTable.

Sort sheet rows by a 1-based column (optionally keep a header row).

Run a VBA macro in an .xlsm file (Windows + Excel + COM).

Read the contents of a text file into a variable.

Write text to a file, overwriting or appending.

List files in a folder matching a pattern (optionally recursive).

Store whether a file exists ('True'/'False').

Move (or copy) a file, or every file matching a wildcard, to a destination folder or path.

Delete a file, or every file matching a wildcard.

Rename a file within its current folder.

Read a file's metadata (name, size, modified time) into a dict.

List the file names directly inside a folder (non-recursive).

Wait until a file appears, up to a timeout.

Compute a file's checksum (MD5/SHA1/SHA256) for verification.

Create a folder (including parents) if it does not exist.

Store whether a folder exists ('True'/'False').

List sub-folder paths inside a folder (optionally recursive, with a name pattern).

Move (or copy) a folder to a new location.

Delete a folder and all of its contents.

Rename a folder within its parent directory.

Total size and file/folder counts of a folder (recursive).

Authorize a Gmail/Google account with OAuth 2.0 (Desktop app) and open a connection for sending, reading, labelling and managing mail. Caches a refresh token for silent reconnects.

Send an email (To/Cc/Bcc, subject, plain-text or HTML body and optional file attachments). Stores the new message id.

Reply to an existing message (kept in the same thread, with 'Re:' subject and In-Reply-To headers). Pass a message id from 'Gmail: Search Messages'.

Search the mailbox with Gmail's query syntax (e.g. 'is:unread from:boss@x.com newer_than:7d') and return matches as a DataTable (messageId, threadId, from, to, subject, date, snippet).

Fetch a single message's full content (from, to, subject, date, body and attachment names) into a dictionary variable. Optionally stores just the body text in a second variable.

Create a draft email (not sent) in the Drafts folder. Stores the new draft id.

Save all attachments of a message to a folder. Stores the list of saved file paths.

Mark a message as read or unread (toggles the UNREAD label).

Add one or more labels to a message (by label name or id). Unknown names are created automatically.

Remove one or more labels from a message (by label name or id).

Move a message to Trash (reversible — not a permanent delete).

Create a new label. If it already exists, returns the existing label id. Stores the label id.

List all labels (system and user) as a DataTable (labelId, name, type).

Authorize with OAuth 2.0 (Desktop app) and open a Drive connection. Caches a refresh token for silent reconnects.

Upload a local file to Drive.

Download a Drive file to a local path.

List files/folders as a DataTable. Filter by name, parent folder or type.

Read a file/folder's metadata into a dict.

Copy a file (optionally rename / place in a folder).

Move a file/folder into a different folder.

Permanently delete a file or folder.

Create a folder in Drive.

Grant access to a file/folder (a person, or anyone with the link). Returns the shareable link.

Authorize with OAuth 2.0 (Desktop app) and open a connection. Caches a refresh token so later runs reconnect silently.

Create a new spreadsheet (document).

Delete an entire spreadsheet (document) via Google Drive.

Read rows. Header row supplies column names (DataTable). Optional lookup filters by a column value.

Append rows, mapping DataTable columns onto the sheet header (writes a header first if the sheet is empty).

Update existing rows matched by a column value (no append).

Update rows matched by a column value; append the rest (upsert).

Clear values from a sheet or an A1 range.

Add a new sheet/tab to a spreadsheet.

Delete a sheet/tab from a spreadsheet.

Delete a range of rows or columns from a sheet/tab.

Capture the screen (or a region) to a PNG file.

Locate a template image within a screenshot (or the live screen).

Read text from an image using Tesseract OCR. Tesseract is an optional dependency — install it once from the Dependencies screen (it is not bundled to keep the installer small).

Resize an image to a width/height, optionally keeping aspect ratio.

Crop a rectangular region from an image.

Rotate an image by an angle and/or flip it horizontally or vertically.

Convert an image to another format (PNG, JPEG, WEBP, BMP, GIF, TIFF).

Re-save an image at a lower quality / optimized size to shrink the file.

Adjust brightness, contrast, color saturation and sharpness (1.0 = unchanged).

Convert an image to black & white (grayscale).

Draw a text watermark/label onto an image at a chosen corner.

Read an image's width, height, format and file size.

Type text using the keyboard into the focused window.

Press a key combination such as Ctrl+C.

Press a sequence of individual keys, one after another.

Click the mouse at screen coordinates.

Move the mouse cursor to screen coordinates.

Show a modern dialog that asks the user to fill in one or more fields (text, dropdown, radio, etc.) and stores each answer in a variable.

Ask the user to pick a file; store its path.

Ask the user to pick a folder; store its path.

Create a new list, optionally pre-filled with values.

Append an item to a list.

Read an item from a list by index into a variable.

Count the items in a list.

Replace an item in a list at a given index.

Remove an item by index, or the first item equal to a value.

Copy a list into a new list variable.

Join two lists into a new list.

Reverse the order of a list in place.

Sort a list ascending or descending, in place.

Serialize a list to a JSON string and/or a .json file.

Keep only the list items that match a condition; store the matching items as a new list (the source list is unchanged).

Remove duplicate items from a list, preserving first-seen order; store the result as a new list.

Take the first or last N items of a list into a new list.

Compute an aggregate over a list's numeric items (Sum / Average / Min / Max) or Count, and store the result.

Store a current-index pointer for a list (as '<name>.Index').

A container that repeats the commands placed inside it. Pick how it iterates: a fixed count, each item of a list, a while-condition, or continuously until Exit Loop.

A colourful sticky note for documenting/grouping the canvas. Purely visual — it never runs and never appears in the log.

Display a message to the user (a dialog box in UI mode).

Pause once, here, for a number of milliseconds, then continue. (For a delay applied automatically before every command, use Set Engine Delay instead.)

Copy text to the system clipboard.

Read text from the system clipboard into a variable.

Clear the system clipboard.

Encrypt or decrypt text with a password (AES via Fernet).

Send an email (optionally with attachments) via SMTP.

Ping a host and store whether it is reachable ('True'/'False').

Append a timestamped message to a local log file.

Compose and send an email via Outlook.

Read recent inbox emails (subject/sender) into a DataTable.

Forward the most recent email matching a subject.

Reply to the most recent email matching a subject.

Move the most recent matching email to a named folder.

Delete the most recent email matching a subject.

Launch a program, optionally waiting for it to finish.

Terminate all processes with the given name.

Run a script file (.bat/.sh/.py/...) and capture its output.

Run a PowerShell script (inline) or a .ps1 file and capture its output.

Run a cmd.exe command (inline) or a .bat/.cmd file and capture its output.

Compile and run Java source code directly. Paste a class with a 'public static void main' method; it is compiled with javac and run with java, and its stdout is captured. (To run a .jar instead, use Run CMD: 'java -jar file.jar'.)

Execute a Python snippet. Variables are supported two ways: {{{name}}} markers are replaced before running, and the live 'variables' dict is also available to read/write. Set 'result' to store an output.

Authorize a YouTube/Google account with OAuth 2.0 (Desktop app) and open a connection for uploading, commenting and liking. Caches a refresh token for silent reconnects.

Upload a video to YouTube with full metadata: title, description, tags, category, privacy, scheduled publish time, thumbnail and optional playlist. Stores the new video id/URL.

Open a Facebook connection from a Page or User access token. Validates the token against the Graph API. Get a Page token from the Graph API Explorer / your app with the right scopes.

Upload a video to a Facebook Page (or profile) with a title and description. Stores the new video id.

Open a TikTok connection from an access token (TikTok for Developers, Content Posting API). Validates it against the user info endpoint.

Upload a video to TikTok via the Content Posting API. 'Inbox (draft)' uploads to the creator's drafts (works for un-audited apps); 'Direct post' publishes with a caption + privacy. Chunked resumable upload.

List top-level comments on a video as a DataTable (commentId, author, text, likeCount, replyCount). Use the commentId with 'YouTube: Reply to Comment'.

Post a new top-level comment on a video.

Reply to an existing comment (use a commentId from 'YouTube: List Comments').

Rate a video as like / dislike, or remove your rating.

List comments on a post/video object as a DataTable (commentId, author, text, likeCount).

Post a comment on a post/video, or reply to a comment by using that comment's id as the object id.

Like a post, video or comment.

Get or set an environment variable.

Read an operating-system property into a variable.

Lock, log off, restart or shut down the computer.

Pre-load a workflow .json file into a named slot so Run Task can call it repeatedly without re-reading the file.

Run another workflow as a sub-step. It shares the current variables, so values set inside flow back to the caller. Give either a pre-loaded slot name or a file path.

Signal a running sub-task to stop at its next step (cooperative — the sub-workflow decides when to check).

Free a slot previously filled by Load Task. Optional — slots are released automatically when the workflow ends.

Keep only the items matching a condition.

Sort items ascending or descending.

Combine a field across items (sum, average, count, ...).

Remove duplicate items (optionally by a field).

Keep only the first or last N items.

Append two lists into one.

Extract one field from a list of objects (or DataTable) into a flat list of values.

Start point for a workflow run started by the user.

Declare a schedule (cron or interval) for a runner to honour.

Declare an HTTP webhook path/method for a runner to host.

Start the workflow on a global keyboard/mouse hotkey (e.g. Ctrl+Alt+R).

Start the workflow when files in a folder are created or changed.

Start the workflow automatically when the app launches.

Find a control and click it, set its text, or read its text.

Store whether a control exists in the window.

Get a control's enabled/visible status as a dict.

Read a control's name into a variable.

Read a control's text into a variable.

Read a cell value from a grid/table control by row and column.

Type text into an editable control.

Click a control.

Left-click a control.

Double-click a control.

Right-click a control.

Create or update a variable with a value.

Set several variables at once — one 'name = value' assignment per line (n8n's Set node). Variables and arithmetic are supported in each value.

Apply a string operation to a variable and store the result.

Evaluate an arithmetic expression and store the result.

Append a value to an existing variable (text or list).

Evaluate a variable's value as an expression and store it back.

Copy one variable's value into another.

Cut a segment from a video by start time + end (or duration).

Join multiple videos into one (re-encoded for safety).

Scale a video. Leave Height blank to keep the aspect ratio.

Crop a rectangle (width x height at x,y from top-left).

Rotate 90/180/270° or flip horizontally/vertically.

Speed up or slow down (e.g. 2 = 2× faster, 0.5 = half speed).

Extract the audio track to MP3 / AAC / WAV.

Replace the video's audio with an audio file (e.g. background music). Trims to the shorter stream.

Scale the audio volume (e.g. 1.5 = louder, 0.5 = quieter).

Remove the audio track.

Overlay an image (logo/watermark) on the video. Supports a margin from the edges, scaling the logo, and an optional animation that walks it clockwise around the four corners.

Burn a text caption onto the video (drawtext).

Add a fade-in at the start and/or fade-out at the end (video + audio).

Re-encode to another container/format with optional quality.

Make a high-quality GIF from a clip (palette-based).

Save a single frame at a time as an image.

A right-to-left marquee text strip near the bottom of the video, with an optional background box and a pause between passes.

Overlay a short effect video on top of a base video, looping the effect until the base ends. Choose how the effect is fitted and how transparent it is.

Build a video from a still image with a looped overlay effect on top, set to one or more music files concatenated. The video lasts as long as the music.

Join two videos with a transition (xfade for video, acrossfade for audio). Both inputs should share the same resolution and frame rate for best results.

Join every video in a folder, in filename order, each pair separated by a transition. Pick one transition or 'Random' to vary it per join. Clips should share resolution/fps.

Tweak brightness, contrast, saturation and gamma (eq filter).

Turn a landscape video into a vertical/square/portrait clip with a blurred copy of itself filling the background — the popular social-media format.

Repeat the whole video N times (no re-encode).

Extract an embedded subtitle track from a video to a standard .srt file you can translate. Fails if the video has no subtitle track (burned-in subtitles are pixels, not text — use 'Cover Subtitle (Glass Blur)' for those).

Add a (translated) .srt subtitle into a video — either as a soft track you can toggle, or burned permanently into the picture.

Cover burned-in subtitle text (e.g. Chinese) with a frosted-glass blurred rectangle — useful before adding your own translated subtitles. Choose the bottom band or a custom region.

Launch a browser session as a named instance. Pick a saved anti-detect Profile (from the Profiles screen) to launch with its browser/fingerprint/proxy, or configure it manually with an optional stealth (anti-bot-detection) mode.

Navigate the browser to a URL.

Read an element's text into a variable.

Type text into an input element.

Perform an action (click, hover, ...) on an element.

Wait until an element appears.

Press a keyboard key on an element, or on the page if no selector is given (e.g. Enter, Tab, Escape).

Set the file(s) of a file <input> element.

Go back to the previous page.

Go forward to the next page.

Reload the current page.

Close the browser session and free resources.

Store whether a named browser instance exists.

Get the current URL and page title as a dict.

Set the viewport width and height.

Run JavaScript and store the result.

Save a screenshot of the current page (full page optional).

Switch the active tab by 0-based index.

Switch element scope into an iframe (blank selector = main).

Clear the value of an input element.

Read an element's attribute into a variable.

Count elements matching the selector.

Store the text of every matching element as a list.

Store the option values of a <select> element as a list.

Select an option in a <select> element by value or label.

Click, double-click, right-click or hover an element.

Bring a window to the foreground.

Close a window.

Move a window to screen coordinates.

Resize a window.

Maximize, minimize or restore a window.

Get a window's state (maximized/minimized/normal).

Get a window's position and size as a dict.

List the titles of all open windows.

Store whether a window with the title exists.

Wait until a window with the title appears.

Create a new empty document as a named instance.

Open an existing .docx file as a named instance.

Add a paragraph of text with an optional style and alignment.

Render Markdown (text or .md file) and append it to the document — headings, lists, tables, code, bold/italic.

Save the document to a specific path.

Read all paragraph text from the document into a variable.

Save the document to its current path.

Close a document instance and discard it from memory.

Store whether a named Word instance exists.

Insert a page break (start the following content on a new page).

Insert an image into the document.

Insert a DataTable variable as a Word table.

Replace all occurrences of text in the document.

Export the saved document to PDF (requires Microsoft Word).