Documentation

Complete guide for building AI workflows and publishing them as agents. Whether you're automating your own tasks or sharing your creations with the community.

What is Neuron Workflows?

Neuron Workflows is a visual AI workflow builder that lets you create powerful automated processes without writing code. Build complex AI pipelines by connecting neurons – intelligent processing units that can read, process, and generate content using advanced AI models.

AI-Powered

Leverage state-of-the-art AI models for text, image, and audio processing.

Visual Builder

Drag, drop, and connect. No coding required to build complex workflows.

Automation

Schedule workflows, trigger on events, and integrate with external services.

For Workflow Builders

Create, run, and automate AI workflows. Install ready-made agents from the community store. Track your credit usage.

Getting Started
For Creators

Publish your workflows as agents in the community store. Earn credits when others install and run your agents.

Creator Guide

Getting Started

1

Create Your First Workflow

From the dashboard, click "Blank Workflow" to open the visual canvas and build from scratch, or use "Describe to Create" to let AI generate a complete workflow from a plain-English description.

2

Add Your First Neuron

Click the "+" button in the sidebar's Neurons tab to add a neuron to the canvas. Each neuron is an independent AI processing unit that is automatically positioned.

3

Configure the Neuron

Each neuron has labeled folder tabs (START, GET, DO, MAKE, SEND) along the top. Click the gear icon on any tab or click directly on a section to configure inputs, AI instructions, output type, and destinations.

4

Run and Iterate

Click the play button (START) to run the neuron. Use "Play" for a single neuron or "Play All" to run an entire chain of connected neurons. Review outputs and refine as needed.

Describe to Create

Not sure where to start? Describe what you want in plain English and AI will generate a complete workflow for you. Be specific about your inputs, outputs, and purpose for the best results.

Example prompts:
• “Summarise PDF reports and email me a bullet-point summary each morning”
• “Take customer feedback text, detect sentiment, and route negative reviews to my support inbox”
• “Translate uploaded audio files to English text and save to a table”

After generation, review each neuron's configuration, test the workflow, and refine as needed.

Understanding Neurons

A Neuron is the fundamental building block of your workflow. Think of it as an intelligent worker that can:

  • Accept various types of input data (text, PDFs, images, audio, audio recordings, tables) with optional input queues for batch processing
  • Process that data using AI according to your instructions
  • Generate outputs in different formats (text, images, audio, tables, or decision-based conditional branching)
  • Send results to other neurons via bridges, email, Telegram, API webhooks, or store to table documents

Neuron Architecture

Each neuron has five main components shown as folder tabs along the top: START (triggers), GET (inputs), DO (AI processing), MAKE (outputs), and SEND (destinations). Each tab includes a gear icon to open its configuration panel.

Workflow Canvas

The canvas is your visual workspace where you design and build workflows. It provides an infinite, zoomable area to organise your neurons and connections.

Navigation

  • Pan: Click and drag on empty space
  • Zoom: Scroll or use the bottom-right zoom controls
  • Mode: Switch between mouse and trackpad interaction modes
  • Reset: Click reset button to centre the view

Sidebar (3 Tabs)

  • Neurons: View neurons organised in flow groups, run/stop, rename, and reorder via drag-and-drop
  • Docs: Create and manage reusable documents (tables, PDFs) shared across neurons
  • Help: Access interactive guides and documentation

Neuron Components

START — Trigger

The START node triggers the neuron to execute. Use 'Play' to run just this neuron, or 'Play All' to execute the entire connected chain. You can also schedule automated runs with countdown timers.

Use cases: Manual execution, chain execution, scheduled workflows, time-based automation

GET — Input Data

The GET node defines what data enters the neuron. Add text, upload files, connect tables, or link bridges from other neurons. Enable an input queue to batch-process multiple items in sequence.

Text Input
Direct text entry
File Upload
PDFs, images, audio
Audio Recording
Record directly in-app
Tables
Structured data
Bridges
Output from other neurons
Input Queue
Batch process multiple items

DO — AI Instructions

The DO node is where you write instructions for the AI. Tell it what to do with the input data – summarise, analyse, translate, extract information, generate content, or any custom task. The more specific, the better the results.

Example instructions:
  • "Summarise this document in 3 bullet points"
  • "Extract all email addresses and phone numbers"
  • "Translate this text to Spanish"
  • "Generate a professional response to this email"

MAKE — Generate Output

The MAKE node produces the final output. Choose your output type and click the corresponding icon to generate results. Decision output enables conditional branching to different neurons based on AI-evaluated conditions.

Text
Image
Audio
Table
Decision
Decision / Conditional Branching

When you select Decision as your output type, the AI evaluates your instructions and routes the result to different connected neurons based on conditions you define. For example: "If sentiment is negative, route to the support flow; otherwise route to the success flow." Create the branches by connecting SEND bridges from this neuron to the target neurons and labelling each branch condition in the DO instructions.

SEND — Route Output

The SEND node routes outputs to one or more destinations. Enable destinations by clicking them to build multi-step processing chains.

Email
Send to addresses
Telegram
Message via Telegram — click the link in the send node to open the bot, send /start, and the bot replies with your Chat ID to paste in
API Webhook
POST to external APIs — configure URL, method, and auth headers in the gear configurator
Store to Table
Save to table document
Bridges
Link to other neurons

Supported Data Types

Text

Plain text, rich text, and formatted content

PDF Documents

Extract and process PDF content

Images

Process and generate images (JPG, PNG, WebP, GIF)

Audio Files

Transcribe and generate audio (MP3, WAV, M4A)

Audio Recording

Record audio directly in the app

Tables

Structured data input and output (CSV, Excel)

Decision

Conditional branching based on AI output

Bridges

Connect neuron workflows together

Bridges & Connections

Bridges are connections between neurons that pass output from one neuron as input to another. This creates powerful processing chains where data flows through multiple AI operations.

Creating a Bridge

  1. 1. In the first neuron's SEND section, click "Neuron Out" to create a bridge
  2. 2. Select or create a new neuron to connect to
  3. 3. In the second neuron's GET section, select the bridge as an input source
  4. 4. The output from the first neuron will automatically flow to the second

Pro Tip

Use bridges to break complex tasks into smaller steps. For example: Neuron 1 extracts data → Neuron 2 analyses it → Neuron 3 formats the results → Neuron 4 sends via email.

Documents

Documents are reusable data sources that can be shared across multiple neurons in a workflow. Instead of uploading the same file to multiple neurons, upload it once as a document and reference it wherever needed.

Benefits

  • • Reuse files across multiple neurons
  • • Centralised document management
  • • Easier updates — change once, reflects everywhere
  • • Supports tables, PDFs, and more

How to Use

  • • Click "Add" in the Docs sidebar tab
  • • Upload or create your document
  • • Reference it in any neuron's GET node
  • • Update once, changes reflect everywhere

Scheduling & Automation

Automate your workflows to run on a schedule without manual intervention. Scheduling is configured from the START node — click the timer icon on any neuron's START section to open the scheduling panel.

Countdown Timer

Set a duration (e.g. 2 hours, 30 minutes) and the neuron will execute once that time elapses. Useful for deferred, one-off tasks.

Scheduled Date & Time

Pick a specific date and time for a one-off execution. The workflow will fire automatically at that moment.

Recurring Schedule

Set a repeating interval so the workflow runs on a regular cadence — hourly, daily, weekly, or a custom interval.

Auto-Cascade Mode

When Auto Mode is enabled in a workflow, downstream neurons run automatically as soon as their upstream neuron completes. Combined with scheduling, this means a single trigger on your first neuron can cascade through an entire multi-step pipeline without any manual intervention.

Monitoring Scheduled Runs

All scheduled executions appear in your Usage page with timestamps, credit costs, and result details. Use this to verify your schedule is running correctly and to review outputs.

Agent Store

The Agent Store is a marketplace of pre-built AI workflow agents published by the community. Browse, install, and run agents without building anything from scratch.

Finding Agents

  • • Browse by category (Productivity, Content, Data Analysis, etc.)
  • • Search by name, description, or tags
  • • View ratings, install counts, and version history
  • • Read the full description and feature list before installing

Installing an Agent

  • • Click an agent card to see full details
  • • Click "Add to My Agents" to install
  • • Choose Reference or Copy (see below)
  • • Find the installed agent in My Agents

Reference vs Copy — Which Should You Choose?

Reference
  • • Stays linked to the creator's original agent
  • Automatically gets updates when the creator publishes new versions
  • • Cannot be edited or modified
  • • Best when you want the latest improvements without thinking about it
Copy
  • • Creates an independent clone of the agent
  • You can open and modify it in the canvas
  • • Will not receive creator updates
  • • Best when you want to customise the agent for your specific needs

My Agents

My Agents is your personal collection of installed agents. Run them on demand, schedule automated executions, and review past results — all without opening the canvas.

1
Select an agentClick any agent in the list on the left to open its execution panel.
2
Fill in inputsThe creator has defined what inputs the agent needs. Fill in the required fields — these may be text prompts, file uploads, or other data specific to that agent.
3
Configure destinations (optional)Optionally set up where results should be sent: email, Telegram, or a webhook. The agent itself may already have destinations configured.
4
RunClick "Run" to execute the agent. Results appear below the inputs immediately after processing.

Scheduling Agent Runs

Click the timer icon on the agent execution panel to schedule a one-off or recurring run. Scheduled agents execute automatically and results appear in the run history.

Run History

Scroll down in the agent panel to see all past executions with inputs, outputs, timestamps, and credit costs. Use this to review results and understand agent performance.

Credits & Usage

Credits are the currency for all AI operations in Neuron Workflows. Every time a neuron runs an AI process — generating text, analysing an image, transcribing audio — a small number of credits is consumed.

What Consumes Credits

  • • AI text generation (DO node processing)
  • • Image generation (MAKE → Image)
  • • Audio transcription and text-to-speech
  • • Running agents installed from the store
  • • Each neuron in a chain consumes credits separately

Managing Your Credits

  • • Your credit balance is shown in the navigation bar
  • • Add credits via your profile menu → Add Credits
  • • Credit cost varies by AI model and operation type
  • • Larger inputs / outputs consume more credits

Usage Page

The Usage page gives you a complete log of every AI transaction. Filter by date range, AI model, or search terms to understand exactly where your credits are going.

  • Summary cards: total transactions, credits consumed, and average cost per transaction for the selected period
  • Records table: each row is one AI operation with model, neuron ID, timestamp, and cost
  • Expand a row: see the full cost breakdown — input tokens, output tokens, and per-unit pricing

Best Practices

Clear Instructions

Write specific, detailed AI instructions in the DO node. The more context you provide, the better the results.

Good: "Extract the invoice number, date, and total amount. Format as: Invoice #X - $Y on Z"

Modular Design

Break complex workflows into smaller neurons. Each neuron should handle one specific task.

Better to have 3 neurons (extract → process → format) than one doing everything.

Test Iteratively

Run neurons individually to test each step before connecting them. This makes debugging easier.

Verify outputs at each stage to ensure data flows correctly through your workflow.

Use Documents

For files used by multiple neurons, create documents instead of uploading repeatedly.

Saves time and makes workflow maintenance much easier.

Workflow Examples

Document Summariser

Upload a PDF → Extract text → Summarise key points → Send summary via email or Telegram

PDF ProcessingText GenerationEmailTelegram

Content Generator

Input topic → Generate blog post → Create social media snippets → Generate cover image

Content CreationImage GenerationMulti-output

Data Extraction Pipeline

Upload invoices → Extract data → Populate table → Send to accounting system via webhook

Data ExtractionTablesWebhooks

Multi-Language Support Bot

Receive message → Detect language → Translate to English → Generate response → Translate back → Send

TranslationChatbotBridges

Lead Qualification Router

Receive inquiry → Analyse intent and budget → Decision branch → Route high-value leads to sales email, others to auto-response

Decision BranchingEmailConditional Logic

Batch Document Processor

Queue multiple PDFs in a single input → Process each sequentially → Extract data → Store results to a table document

Input QueueBatch ProcessingTables

Troubleshooting & FAQ

Common Issues

Workflow run produces empty output

Refine your DO instructions — be more specific about the task. Verify that input data is loaded correctly in the GET node before running. Check if the selected output type matches what you're asking the AI to produce.

File upload is rejected or not processed

Confirm the file format is supported (PDFs, JPG/PNG/WebP images, MP3/WAV/M4A audio, CSV/XLSX tables). Check the file isn't corrupt or password-protected. Large files may take longer to process.

Telegram messages not arriving

Ensure you completed the full bot setup: click the Telegram link in the SEND node, open the bot in Telegram, send /start, and paste the Chat ID it replies with into the configuration.

Webhook endpoint not receiving data

Verify the webhook URL is publicly accessible (not localhost). Check the HTTP method and any required auth headers in the gear configurator. Use a tool like webhook.site to test the request payload first.

Credits used but no result produced

AI processing consumes credits even when a run fails or returns no output. Check the Usage page for the specific transaction. Refine your instructions and retry.

Cannot publish an agent

Your workflow must have exactly one connected flow group before it can be published. Ensure all neurons are connected with bridges and there are no isolated neurons on the canvas.

Frequently Asked Questions

Creator Guide

Build. Publish. Earn.

Once you've built a workflow you're proud of, you can publish it to the Agent Store as a reusable agent. Others install and run your agent — and you earn credits every time they do. This section explains everything a creator needs to know.

Publishing Your Agent

Any workflow you build can be published to the Agent Store as a reusable agent. Other users can then install it, run it with their own inputs, and you earn credits each time they do.

Before You Publish — Prerequisites

  • • Your workflow must contain exactly one connected flow group — all neurons linked by bridges into a single chain
  • • Test your workflow thoroughly — published agents are run by real users
  • • Ensure your DO instructions are robust and handle edge cases gracefully
  • • Decide which inputs you want users to provide vs which are fixed

Publishing Steps

1
Open the Publish dialog

Click the "Publish" button in the canvas toolbar (accessible via the Advanced Features tour). This opens the agent publishing dialog.

2
Fill in agent details

Provide a clear name, a detailed description, relevant category, searchable tags, and optionally a logo image. Good descriptions directly impact your agent's discoverability in the store.

3
Configure exposed inputs

Choose which inputs users will fill in when they run your agent. These become the form fields displayed in My Agents. Fixed inputs (like a processing template) stay hidden from users; dynamic inputs (like user-provided text) are exposed.

4
Set pricing

Choose your pricing model: free (users can install and run without payment), one-time purchase (users pay credits to install), or royalty (you earn a share of credits consumed each time the agent runs). See the Pricing & Earnings section for details.

5
Publish

Submit your agent. Once published, it appears in the Agent Store and is discoverable by all users. You can update or unpublish it at any time.

Writing a Good Description

  • • Start with what the agent does in one sentence
  • • Describe the inputs it expects and outputs it produces
  • • List specific use cases or scenarios
  • • Mention any limitations or requirements

Updating a Published Agent

  • • Modify the workflow and republish to create a new version
  • Reference users automatically receive your update
  • Copy users remain on the version they copied
  • • Maintain backward compatibility where possible

Pricing & Earnings

As a creator, you earn credits when users interact with your published agents. Neuron Workflows supports two pricing models, and deducts a platform commission from each transaction — you keep the rest.

One-Time Purchase

Users pay a fixed number of credits to install your agent. You receive a one-time payment when they purchase.

  • • Good for well-defined, standalone tools
  • • Appears as Purchase in your earnings history
  • • Platform commission is deducted at point of purchase

Royalty (Per Run)

You earn a share of the credits consumed each time a user runs your agent. Income grows with usage.

  • • Good for agents users will run repeatedly
  • • Appears as Royalty in your earnings history
  • • Platform commission is deducted per run

How the Commission Works

Neuron Workflows deducts a platform commission from each transaction before adding the remainder to your creator wallet. The commission amount is shown in your earnings history table next to each record so you can see exactly what was deducted. Your creator earnings (the amount after commission) are what accumulate in your Available Balance.

Maximising Earnings

Agents with clear descriptions, relevant tags, and good ratings rank higher in the store. Consider the royalty model for utility agents that users run daily, and the purchase model for specialised one-off tools.

Creator Dashboard

The Creator Dashboard is your earnings control centre. Access it from your profile menu or the Agent Store page. It shows all earnings from your published agents in real time.

Available Balance

Credits earned and ready to withdraw. This increases as users purchase or run your agents (after commission deduction).

Lifetime Earnings

Total credits earned since you started creating, including both one-time purchases and all royalty payments.

Total Withdrawn

The cumulative amount you have previously withdrawn from your creator wallet to your main account balance.

Purchase Earnings

Credits earned specifically from one-time agent purchases, separate from royalty income.

Earnings History Table

Every transaction that generated earnings appears in the history table. Each row shows:

Agent
Name, logo, and version
Type
Purchase or Royalty
Your Earnings
After platform commission
Date
Timestamp of transaction

Withdrawing Your Credits

When you have an Available Balance, click the Withdraw button at the top of the Creator Dashboard. Credits are transferred from your creator wallet to your main account balance, where you can use them to power your own workflows. A minimum balance may be required to initiate a withdrawal.

Ready to Build Your First Workflow?

Start creating powerful AI workflows today. No coding required, unlimited possibilities.

Ready to Build Your First Workflow?

Start creating powerful AI workflows today. No coding required, unlimited possibilities.