Introduction to Nodes

Learn the four main node groups in Draft & Goal — Inputs, AI, Integrations, and Tools — and how to choose the right node for your workflow

What is a node?#

A node is one step in a workflow: it takes data in, does one job, and passes results to the next step. You build Apps by connecting nodes on the canvas.

For ports, variables, and configuration basics, see Nodes (fundamentals). This page focuses on how the library is organized and how to pick the right node.

The four main groups#

The Node Library mirrors four families. Start here before searching by name.

Group	Role in the workflow	Typical question it answers
Inputs	Collect data when the App runs (or embed fixed values)	“What must the user provide?”
AI	Generate, transform, or analyze content with models	“Do I need a language or generative model?”
Integrations	Read from or write to an external product	“Is the data already in Sheets, Search Console, HubSpot…?”
Tools	Transform data, control flow, or run custom logic	“Do I need to filter a list, loop, branch, or clean HTML?”

A solid workflow almost always combines several groups: Input → Integration or AI → Tools → Integration (write) — with Tools in the middle whenever you reshape data or control the path.

Inputs#

Inputs define the App form (and fixed values) before a run starts.

Subgroup	When to use	Examples
Basic Inputs	Free-form fields filled by the runner	Short Text, Paragraph, URL, Date, File, Select, Brand Kit Variables
Integration Inputs	Let the runner choose a resource from a connected account	Google Drive, Analytics Property, Search Property, Ads Campaigns
Static Inputs	Values that never change between runs (no form field)	Static Text, Static File, Static Image, Static Video, Static Constants

Do not confuse Inputs → Integration Inputs (pickers on the launch form) with Integrations nodes (fetch or push data during the run). Same products, different moment in the lifecycle.

→ Inputs overview

AI#

AI nodes call models: text, structured analysis, images, or video.

Need	Prefer
Write or rewrite text from a prompt	LLM
Multi-step reasoning with tools	AI Agent
Ask questions about a CSV / spreadsheet	Data Analyzer
Create or transform images / video	Text to Image, Image to Image, Upscale Image, Image to Video, …

Use AI when the step requires judgment, generation, or understanding — not when you only need a deterministic transform (prefer Tools) or a known API (prefer Integrations).

→ AI overview

Integrations#

Integrations talk to external systems during execution: scrape a page, read Sheets, pull Search Console, update HubSpot, publish to WordPress, call SEO APIs, and more.

Ask:

1. Is the source or destination a named product? → look under that provider (Google, HubSpot, Semrush…).
2. Do I only need the page HTML? → Web Scraper.
3. Is there no dedicated node? → API Connector or Code Block under Tools.

Most integration nodes need an org or personal connection under Builder → Integrations before they run.

→ Integrations overview

Tools#

Tools stay inside Draft & Goal: they reshape data and control how the graph runs.

Family	Use when you need to…	Examples
List operations	Split, filter, merge, count, or loop over items	Loop, Filter List, Merge Lists, Count List Items
Data / text transform	Extract JSON fields, clean HTML, merge templates	JSON Path Extractor, HTML Cleaner, Merge With Template, Find & Replace
Flow control	Branch, stop on error, wait for a human	Conditional, Stop Node, Human in the Loop
Advanced	Custom HTTP or Python	API Connector, Code Block
Media / output helpers	Crop images, send email, process video	Change Aspect Ratio, Email Sender, Video Merger

→ Tools overview

How to find the right node#

Use this sequence when you are stuck in the library.

1. Name the job in one sentence#

Examples: “User pastes a URL”, “Pull GSC queries”, “Rewrite the title with AI”, “Keep only rows where volume > 100”, “Write the result to Sheets”.

2. Map the sentence to a group#

flowchart TD
  start[What does this step do?]
  start --> q1{Needs a human value at launch?}
  q1 -->|Yes| inputs[Inputs]
  q1 -->|No| q2{Calls an external product or URL?}
  q2 -->|Yes product API| integ[Integrations]
  q2 -->|Yes generic page| scrape[Web Scraper]
  q2 -->|No| q3{Needs generative AI?}
  q3 -->|Yes| ai[AI]
  q3 -->|No| tools[Tools]

3. Search, then open the category#

1. Open the Node Library on the canvas.
2. Use the search bar with a verb or product name (sheet, loop, scrape, llm).
3. If nothing fits, open the group from the table above and scan the overview pages.
4. Prefer a dedicated Integration over Code Block when one exists — less maintenance, clearer credentials.

4. Common confusions#

You might think…	Prefer instead…
“I need Google Sheets as an input field”	Inputs → Google Drive (or a text/URL field) to select; Sheets Reader / Writer under Integrations to read or write data
“AI can filter my list”	Filter List or JSON Path Extractor for deterministic filters; AI only if the rule is fuzzy
“Loop and LLM are the same”	Loop iterates items; LLM generates text — often Loop → LLM together
“Static Text vs Short Text”	Static = fixed in the workflow; Short Text = filled each run on the App form

5. Check the category overview#

Each group has an overview that lists every documented node with a one-line purpose. Start there when you explore a new area:

Typical chain#

Most production Apps follow a pattern like this:

flowchart LR
  A[Input URL] --> B[Web Scraper]
  B --> C[HTML to Markdown]
  C --> D[LLM]
  D --> E[Human in the Loop]
  E --> F[Sheets Writer]

1. Input — what the runner provides
2. Integration — fetch external data
3. Tool — clean or structure
4. AI — generate or analyze
5. Tool — approval or branching
6. Integration — write the result back

Next steps#