Build automation from scratch

This is a full walkthrough: create a recording → choose an agent → create an automation → finish development in chat → ask the agent to adjust the create form → run it for real.

Still using Browser as the example. Before you start:

• Finish Quick start (client installed and signed in; download first if needed)
• Have a fingerprint browser you can start (Platform browser is fine)
• Before recording, turn off top-right Tasks (task execution master switch; some recording copy says “task running”). If it is still on, starting a recording may show Please turn off task running first

---

Goal of this demo

Build a browser automation that “opens a given page → performs one business action”, and let operators fill parameters such as Target URL when creating a task group.

Each step matches real client menus—follow along in the app.

---

Step 1: Create a recording

A recording captures how a human operates. The agent uses it as a reference; it is not the final executable script by itself.

1. Open Recording in the left sidebar.
2. Click + next to the sidebar title to open New recording.
3. Fill in Basic info, then click Create:
   • Name: e.g. Demo - open page and act
   • Device type: Browser
   • Social type: pick the closest match
   • Recording device: your fingerprint browser (device used for cast demo)
4. In the recording detail, click Add track (one track = one scenario; beginners need only one).
   • Set Track name and optional Sample entry URL
5. Click Record this track to enter the cast recording page:
   • Operate the real flow in the cast view
   • Add snapshots, step titles, and instructions (“click where / type what”)
6. Click Finish recording (or Save draft and exit first). Status becomes Completed.

Tip: split complex flows into multiple parallel tracks. One track is enough for this demo. Only Completed recordings can be selected in the next step.

---

Step 2: Choose an agent and create automation

When you pick an RPA expert template, you go straight into the Create automation wizard (there is no separate “initialize workspace” step—the workspace is handled automatically).

1. Open Agents → New agent (or Recommended agents).
2. On Select agent template, choose Browser automation expert (Android / iOS experts for those platforms; System assistant for Q&A only).
3. Complete the Create automation wizard in three steps:

A. Select recording

• Pick a Completed recording that matches the device type and is not yet bound to another automation.
• If the list is empty, finish a recording in the Recording module first.

B. Project ID

• Enter a Project ID: lowercase letters and hyphens, e.g. demo-open-page.
• Project name, Social type, and Recording device are auto-filled from the recording (read-only). That recorded device is your debug device—make sure it is started in Browser.

C. Enhancements (optional)

• Defaults are already on; skip unless you need extras.
• Click Start conversation to open the project’s agent chat.

The project enters development mode and later appears under Automation → My automations (batch runs make sense after chat-driven development finishes).

---

Step 3: Finish the automation in chat

Drive development in natural language (Development mode), for example:

Please develop the automation from the recording. The debug device is already started—write and trial-run as you go.

Then add constraints such as:

• “Step 2 search text should come from the task-group form, not hard-coded”
• “If a login dialog appears, pause and notify me”

The agent writes/maintains the script from recording steps and can Trial-run script. Watch Device casting in the side panel.

When satisfied, confirm the project appears under Automation → My automations and can create task groups.

---

Step 4: Ask the agent to adjust the create form

Fields shown when creating a task group come from project config. Change them in chat, for example:

Please add task-group form fields:

1. Target URL (required, text)
2. Notes (optional, text)
   And read the hard-coded URL from “Target URL” instead.

Then:

1. Open Create preview in the chat side panel, click Refresh, and review the form.
2. Keep iterating: “Make Notes multiline”, “Add an https hint for the URL”, etc.
3. Trial-run script again to confirm the script reads the new fields.

---

Step 5: Run for real

1. Open Automation → My automations.
2. Click the project card body (not the expert button) to open create task group.
3. Fill in the task group name, plus Target URL, Notes, etc., then click Create.
4. Confirm schedule and counts in Confirm submit.
5. Turn on top-right Tasks when prompted.
6. Open the group under Task group management to watch progress and results.

Notes:

• Card Browser automation expert → continue Development mode chat. For a failed task group, use Task group management → Investigate and fix (Troubleshooting mode).
• Production task-group runs usually use almost no tokens; chatting for changes or fixes is what consumes tokens.

---

Demo checklist

Step	What you do in the client	Outcome
1	Recording → create → add track → finish	Completed reference recording
2	Agents → Browser automation expert → Create automation wizard	Pick recording + project ID → start chat
3	Chat: “Develop from recording and trial-run script”	Script runs
4	Chat: adjust create form + Create preview	Task-group form ready
5	My automations → card → confirm submit → enable Tasks	Production batch run

---

Related

• Quick start
• Devices & automation
• Workflows & content
• Case Studies · Download · Register · Sign in