Platform
Deployment, authorization, and governance in one system
The platform provides desktop operation, browser operation, vision, workflow orchestration, and sandbox permissions so every execution role can be authorized, limited, observed, paused, and confirmed by humans at key points.
Execution client
A digital employee client with infrastructure built in
To enter business systems, digital employees must operate computers, browsers, and interfaces like people, while staying within controlled permissions.
Desktop operation
Operate desktop applications like an employee to download, organize, and export.
Browser operation
Log into admin panels, read pages, fill forms, and cover systems without APIs.
Vision analysis
Recognize product images, prints, labels, tickets, and UI content for decisions.
Sandbox permissions
Execute in restricted environments with grantable, revocable, auditable permissions.
Tool calling
Dispatch APIs, scripts, and internal systems to turn judgment into execution.
Workflow orchestration
Compose multi-step tasks into reusable flows across systems.
Two equally important concepts
Workflows and employees are the platform core
A workflow defines how work proceeds step by step. An employee defines who acts with goals, permissions, and tools. The platform carries both.
Workflow
Reusable multi-step processes: trigger, read, judge, execute, approve, and return results across systems.
Execution role
An actor with role, goal, permission boundary, and toolset, loadable into workflows.
Delivery form
Deliver a runnable execution system
From entry to execution and result return, six modules form a deployable and governable digital employee system.
Unified entry
Use Lark Bot as the entry point; employees drive tasks with natural language, files, and approval cards.
Analysis and strategy
Understand intent, analyze data, generate suggestions, and judge risk.
Automation executors
Connect GUI automation, API automation, file processing, and data processing.
Rules and knowledge base
Accumulate violation libraries, campaign strategies, role profiles, and operation rules.
Approval and permission control
Route high-risk actions to human confirmation with audit traces.
Reports and result return
Return execution results to Lark and generate deliverable files.
Different from traditional RPA
Why this is not traditional RPA
Traditional automation mainly replays fixed steps; digital employees understand tasks, analyze context, judge risk, and accumulate experience.
Governance and control
Digital employees must be controllable, auditable, and take-over ready
The platform is designed around permission sandboxing, logs, human confirmation, risk thresholds, exception pause, replay, and role authorization.