Browser Workflows
Read the current page, extract selected text, inspect metadata, collect links and images, fill forms, click controls, wait for page changes, and display results back in the browser.
Agentic WorkFlow Documentation
Agentic WorkFlow
Documentation
Build browser-native workflows that read pages, transform data, call services, and use AI models through a visual node system.
What You Can Build
Section titled “What You Can Build”Data Pipelines
Move items between nodes, map fields, branch with conditions, merge streams, loop through records, format values, and send clean output to APIs or integrations.
AI Systems
Connect chat models, embeddings, vector stores, memory, output parsers, and agent nodes to summarize, classify, answer questions, or reason over page content.
Service Automation
Combine browser context with HTTP requests and integrations such as Google Sheets, Gmail, Notion, Slack, Airtable, and Google Drive.
How Agentic WorkFlow Thinks About Automation
Section titled “How Agentic WorkFlow Thinks About Automation”Every workflow is a small execution graph. A trigger starts the run, each node receives data from previous nodes, and the workflow either updates the browser, transforms data, calls an external service, or asks an AI model to produce an answer.
flowchart LR
Trigger["Trigger"] --> Context["Browser or input context"]
Context --> Action["Node actions"]
Action --> Data["Structured items"]
Data --> Decision{"Branch, merge, or loop"}
Decision --> Output["Output, page update, API call, or AI response"]
The most reliable workflows keep this graph explicit: know what starts the run, what data each node outputs, where the browser state can change, and where failures should stop or branch.
Start With The Right Section
Section titled “Start With The Right Section”First Workflow
New to the app? Start with the quick intro, then build a small workflow that reads a page and returns a visible result.
Key Concepts
Learn the mental models behind workflow lifecycle, execution order, browser context, item linking, data mapping, loops, merges, waiting, and errors.
Node Reference
Use the source-aligned node reference when you need exact node behavior, inputs, outputs, credentials, dependencies, and troubleshooting notes.
Advanced AI
Learn when to use model dependencies, agents, RAG, memory, structured outputs, tool selection, and evaluation patterns.
Node Palette Map
Section titled “Node Palette Map”The node documentation follows the same group order as the app palette:
- Trigger - start a workflow manually, from the browser, or from a page event.
- Lambda - package reusable workflow logic behind clear inputs and outputs.
- In Page Action - read, inspect, display, or interact with the active browser page.
- Flow - branch, wait, merge, loop, and stop workflows.
- Data Transformation - edit fields, pick values, format dates, and download output.
- Core - run code, make HTTP requests, read URLs, and process feeds.
- AI - build model chains, agents, RAG workflows, and reusable AI dependencies.
- Integrations - connect workflow output to external services.
Practical Learning Path
Section titled “Practical Learning Path”Use this order when you are learning or debugging:
- Build a simple browser workflow with Get Selected Text or Get All Text.
- Read Data Structure and Item Linking so mappings make sense.
- Add control flow with If, Merge, Loop, and Wait.
- Add AI only after the data shape is stable. Start with Model Dependencies and Prompting and Structured Outputs.
- When something breaks, use the Troubleshooting Guide to isolate whether the issue is browser state, data mapping, credentials, timing, or model output.
Agentic WorkFlow is easiest to learn when each workflow stays observable: run small sections, inspect node outputs, name important fields clearly, and keep browser actions separated from data processing whenever possible.