Loading projects...
Workflow Builder
How to Use the Workflow Builder
Getting Started
- Click "New Workflow" to create a workflow
- Add recorded processes from the right panel
- Click "Edit" to enter edit mode
- Connect processes to build your automation chain
- Click "Run" to execute the workflow
Adding Processes
- Browse Available Recordings in the right panel
- Click a recording to add it as a node
- Use API Input/Output for webhook integrations
- Use AI Data Node to transform data between steps
Connecting Processes
- Enter Edit mode to manage connections
- Click the link button between nodes to connect them
- Data flows from one process to the next
- Click "Disconnect" to remove a connection
Running Workflows
- Click "Run" to execute all connected processes
- Watch the status bar for execution progress
- Results appear in the Scraped Data tab
- Failed steps show error details for debugging
Tip: Workflows run sequentially. Order matters!
Tip: Use API hooks for external integrations.
Tip: AI Data Nodes can transform between steps.
Workflows
No workflows yet
Click 'New Workflow' to get started
Select a workflow to view
Or create a new one to get started
Available Recordings
No recordings available
Create recordings in the Recorder tab
How Workflows Work
Getting Started: Create a workflow and add recorded processes as nodes.
Connections: Link nodes to chain processes. Data flows from one to the next.
API Hooks: Add API Input nodes for webhooks, Output nodes for responses.
AI Data: AI Data Nodes transform data between processes intelligently.
Auto-Save Settings
No scraped data yet
Run a scraping automation or use the Claude Agent to collect data
Drop files here or click to upload
No files in the vault
Files captured during recordings will appear here
No SOP documents yet
Enable "SOP Recorder" in the Recorder to auto-generate SOPs
Variables
Recorded Script Manager
Process: —
| Label | Event | Path | Input |
|---|
| No steps recorded yet. Click Record to begin. |
🔴 Playwright Codegen is running
Record actions in the Playwright browser window, then click Fetch Script.
// Waiting for steps...
Conditional Click
Click this element only if a condition is met. The automation will check for the condition before executing.
AI Agent Workflow Step
Short label shown in the step list.
Stop after this many actions.
Overall time limit.
Safety Review Before Playback
Please review the steps below before running. The automation will interact with live websites on your behalf.
Refine SOP Labels
Edit the auto-generated labels for each recorded step. These will be saved in your process.
👋 Welcome to Recorder!
Settings
Configure API keys, recording behavior, and extension preferences
Subscription & License
Your current plan and license details
Free Trial Plan
Get started with basic automation
License Key
—
Activated
—
Anthropic API Key
Required for AI-powered features (Claude Agent, Vision fallback). Get your key from Anthropic Console
How Tier 3 Vision API Works
- Activates when DOM selectors (T0) and handler patterns (T1) fail
- Captures screenshot of current viewport
- Sends screenshot + element description to Claude Vision API
- Claude returns coordinates of matching element
- Cost: ~$0.004 per Vision API call (~0.4 cents)
Privacy Note
API key is stored locally in chrome.storage.local and never sent to any server except Anthropic's API.
Recording Behavior
Configure how recordings are captured
When ON, recording attaches to existing tab. When OFF, opens new tab.
Shows modal during tier escalation to prevent accidental clicks.
Allow recording in multiple tabs simultaneously. Experimental.
Captures additional recovery data. Uses ~25x more storage.
Generate Standard Operating Procedure documents while recording.
AI Agent & Logic Engine
Enable the Claude AI agent and select which reasoning engine it uses during recording.
When enabled, the AI sidebar opens when recording starts. Claude assists with complex steps using the selected logic engine.
Select the reasoning engine the Claude agent uses. Takes effect on next recording session.
Developer Options
Debugging and telemetry settings
Enable verbose logging for troubleshooting.
Help improve Auto by sharing anonymous usage stats. No personal data.
Shows raw JSON export in export menu for developer troubleshooting.
AI-powered browser overlay for automated interactions.
Shows Field Mapper in sidebar navigation for CSV field mapping.
Local watcher that lets the website run shell/Python/.bat commands on your PC. Auto-starts at Windows login if you ran INSTALL_STARTUP.bat.
Enter access code for agent integration features.
Canvas
Open a hidden localhost portal that lets Cowork / Claude watch and control this desktop with vision + OS-level click and type. Accessible only with a one-time access key.
Generates an access key for http://127.0.0.1:5055/canvas.
Requires canvas-service/start_canvas.bat to be running.
Export Options
Show or hide export buttons on the Recording toolbar.
Shows the "Export Process" button on the Recorder toolbar to save recordings as .automation files.
Shows the "Export Header CSV" button on the Recorder toolbar to export step headers as CSV.
Shows the "Export JSON" button on the Recorder toolbar to export the full recording as JSON.
Developer Tools
Build, reload, and debug the extension from this dashboard.
Display Options
Service Management
Control the background services that power Auto automation. Use "Setup Auto-Start" so everything launches automatically when you log in — no PowerShell needed.
Python daemon that lets Claude control your PC (mouse, keyboard, screenshots, shell commands).
Node.js server at localhost:8765 — serves this website and relays WebSocket messages.
Registers both services with Windows Task Scheduler so they start silently whenever you log in. Run once — never touch PowerShell again.
Data & Files
Access your scraped data and uploaded files.
About
Version
2.3.0
Build
Internal (All features)
Plan
Free Trial
Tier Path
T0 → T1 → T2 → T3 → T5
Reports & Analytics
Detailed analytics charts coming soon
Automation Triggers
WHEN a condition is met → DO an action automatically
0
0
0
No automation triggers yet
Create a trigger to automate your workflows
API & Webhooks
Test endpoints, manage webhooks, and trigger workflows via API
Endpoint Tester
Response
// Send a request to see the response here
Webhooks
Register webhooks to receive notifications when workflows complete, steps fail, or other events occur.
Request History
Field Mapper
Map CSV columns to recorded step inputs for data-driven testing
Select Process
Upload CSV
Click or drag a CSV file here
Supports .csv and .tsv files
Column Mappings
Map each CSV column to a step input field
Data Preview
Test Runner
Project: —
Select a process to view its steps
Run a test to see results
Test history will appear here
Step execution trace will appear here
Choose Your Plan
Scale your automation with the right plan for your team
Free
$0/mo
Get started with basic automation
- 5 processes
- 50 steps per process
- Basic recording
- Community support
Starter
$19/mo
For individuals and small teams
- 25 processes
- 200 steps per process
- CSV loop mode
- Email support
Pro
$49/mo
Full power automation
- Unlimited processes
- Unlimited steps
- AI Agent & workflows
- API & webhooks
- Priority support
Enterprise
Custom
For large organizations
- Everything in Pro
- SSO & team management
- Custom integrations
- Dedicated support
Profile
User
Not signed in
Current Plan
Free Trial
License Status
Active
Usage Statistics
0
Processes
0
Total Runs
0
Steps Recorded
When should this run?
Choose a trigger type for your automation
Configure Trigger
Trigger Created!
Claude Agent
AI-powered browser automation assistant
Proposed Plan
Confirmation Required
Medium RiskDecision Trace
Generated Content
Ask Claude to automate a task on any website
Select a process above or describe a new task below
Session History
Desktop Agent
AI-powered automation for Claude Desktop & Cowork — auto-approve prompts, submit tasks, or run fully autonomously
Auto-Approve
Watches your screen and automatically clicks "Allow" / "Yes" / "Approve" on Claude permission dialogs. No more babysitting.
Submit Task
Opens Claude Desktop / Cowork, types your prompt, and submits it — then hands control back to you.
Full Agent
Submits your task AND auto-approves all prompts while it runs — fully autonomous end-to-end.
The agent is watching your screen. Move mouse to top-left corner to emergency stop.
Activity Log
No activity yet. Start an agent mode above.