Skip to main content
Blueprint Blueprints are built from nodes, and each node type is designed for a specific kind of work. Choosing the right node type keeps your Blueprint readable and ensures you get the right behavior from the execution engine. The sections below group node types by purpose so you can find the right one quickly.
  • AI / LLM
  • Data
  • Logic
  • Integrations
  • Human
  • Agent
These nodes call AI models to generate, transform, or analyse content. They are the core of most Blueprints. ai-prompt — Text generation The ai-prompt node sends a message to a language model and returns the generated text. Use it any time you need the model to write, summarise, classify, extract, translate, or reason over text. When creating a Blueprint through the API, use ai-prompt as the node type. Example configuration
image_generation — Generate images from text The image_generation node sends a text prompt to an image model (such as DALL·E or Stable Diffusion) and returns a generated image. Use it to create illustrations, product mockups, or visual assets on the fly. Key config fields: prompt, model, size, quality, n (number of images). image_editing — Edit or transform existing images The image_editing node takes an input image and applies model-guided edits based on a text instruction. Use it to retouch photos, apply styles, or modify regions of an image. Key config fields: image (URL or base64), instruction, model, mask. image_variations — Generate variants of an image The image_variations node produces alternative versions of a source image while preserving its overall composition. Use it to explore creative directions or generate multiple options. Key config fields: image, n, size, model. vision — Analyse images with a multimodal model The vision node passes one or more images to a multimodal model alongside a text prompt. Use it to describe images, extract text from screenshots, or answer questions about visual content. Key config fields: images (array of URLs or base64), prompt, model. ocr — Extract text from images The ocr node runs optical character recognition on an image and returns the extracted text. Use it to digitise scanned documents, receipts, or screenshots before feeding the text into an ai-prompt node. Key config fields: image, language. speech_to_text — Transcribe audio The speech_to_text node transcribes an audio file into text. Use it to process voice recordings, meeting audio, or podcast episodes before running downstream analysis. Key config fields: audio (URL or base64), model, language. text_to_speech — Convert text to audio The text_to_speech node synthesises spoken audio from a text string. Use it to build audio summaries, voice interfaces, or accessibility features. Key config fields: text, voice, model, speed. video_analysis — Understand video content The video_analysis node sends a video (or keyframes) to a multimodal model for analysis. Use it to summarise video content, detect events, or extract structured information from recordings. Key config fields: video (URL), prompt, model. data_extraction — Pull structured data from text The data_extraction node instructs a model to extract structured fields from unstructured text and return them as JSON. Use it to parse emails, reports, or any free-form document into a schema you define. Key config fields: input, schema (JSON Schema), model. Data nodes connect your Blueprint to external data sources — APIs, databases, files, and cloud storage. api — Make HTTP requests The api node makes an HTTP request to any REST endpoint and returns the response. Use it to fetch live data, post to third-party services, or call your own backend from within a Blueprint. Example configuration
data — Pass static or transformed data The data node lets you inject a fixed JSON value or transform an existing value into the chain. Use it to provide seed data, reshape an upstream response, or create a reference object other nodes can read. Key config fields: value (any JSON), transform (optional expression). cloud-drive — Read files from cloud storage The cloud-drive node retrieves a file from a connected cloud storage provider (such as Google Drive or Dropbox) and makes its content available downstream. Use it to process documents, spreadsheets, or images stored in your team’s cloud drive. Key config fields: connectionId, fileId, outputFormat (text | base64 | url). Logic nodes control the flow of execution and let you run custom code inside a Blueprint. conditional — Branch on a condition The conditional node evaluates a boolean expression and routes execution down one of two paths. Use it to skip steps, handle error states, or apply different transformations based on the data flowing through the chain. Key config fields: condition (JavaScript expression), truePath, falsePath. function — Run custom JavaScript The function node executes a snippet of JavaScript (or TypeScript) you write directly in the editor. Use it when you need logic that no built-in node covers — data transformation, string manipulation, or complex calculations. Key config fields: code (JavaScript function body), timeout. The function receives the previous node’s output as its first argument and must return a value:
Integration nodes connect your Blueprint to external platforms and protocol servers. mcp — Connect to an MCP server The mcp node connects to a Model Context Protocol server and calls one of its tools. Use it to give your Blueprint access to context from external systems — file systems, databases, search indices, or any MCP-compatible service. Key config fields: connectionId, tool, arguments. webhook — Emit an event to an external URL The webhook node sends an HTTP POST to a URL you specify, typically to notify an external system that a step has completed. Use it to trigger Zapier automations, Slack notifications, or any webhook-compatible service. Key config fields: url, headers, payload. integration — Connect a pre-built integration The integration node uses a saved connection from your integrations library to call a supported third-party service. Use it instead of the api node when Blueprint has a native integration for the service you need, as it handles authentication automatically. Key config fields: integrationId, action, parameters. Human nodes pause execution and require input or approval from a person before the Blueprint continues. human — Collect human input The human node pauses the Blueprint and presents a form to a specified user or role. Execution resumes only after a human submits the requested input. Use it for data collection, content review, or any step where automation alone is not sufficient. Key config fields: prompt (instructions shown to the reviewer), fields (array of form fields), assignee. human_checkpoint — Require approval to continue The human_checkpoint node pauses execution and asks a designated approver to review the current state and either approve or reject. If rejected, execution stops. Use it for compliance gates, quality checks, or spend approvals. Key config fields: message, approvers, timeout. approval_gate — Multi-party approval The approval_gate node requires approval from multiple users or a quorum before proceeding. Use it for high-stakes decisions that need more than one sign-off. Key config fields: approvers, requiredCount, message, timeout. document — Generate a structured document The document node assembles a formatted document (PDF, Markdown, or HTML) from a template and the data flowing through the chain. Use it to produce contracts, reports, proposals, or any structured output that needs to be stored or sent. Key config fields: template, format, data. process_step — Model a business process step The process_step node represents a manual task within a larger business process. It records that a defined step must be completed by a human operator and tracks its completion status. Use it when integrating Blueprints into structured SOPs or audit workflows. Key config fields: name, description, assignee, dueDate. Agent nodes orchestrate autonomous AI agents that can plan, use tools, and complete multi-step goals. agent — Run a single autonomous agent The agent node runs an AI agent that can call tools, reason over results, and iterate until it achieves its goal. Use it when the task cannot be expressed as a fixed sequence of steps and requires dynamic decision-making. Key config fields: goal, model, tools, maxIterations. agent_team — Coordinate multiple agents The agent_team node runs a group of specialised agents that collaborate to complete a shared goal. Use it for complex research, multi-domain analysis, or tasks that benefit from parallel expertise. Key config fields: goal, agents, supervisor, maxIterations. goal — Define a high-level objective The goal node describes a high-level objective and lets the engine decompose it into sub-tasks automatically. Use it as the entry point for agent-driven Blueprints where the exact steps are not predetermined. Key config fields: description, successCriteria, model.