UFO vs Browser Use

UFO² captures Windows desktop screenshots, annotates UI elements with bounding boxes and semantic labels, and executes actions (clicks, text input, keyboard commands) by mapping LLM-generated action descriptions to concrete UI coordinates. The system uses OCR and UI inspection APIs (COM-based Windows Automation Framework) to build a semantic representation of the screen state, enabling the agent to interact with any Windows application without requiring native API bindings or application-specific integrations.

Unique: Combines hierarchical agent architecture (Host Agent for window/app selection + App Agent for UI interaction) with multi-modal prompting (screenshots + OCR + UI annotations) to enable agents to reason about desktop state and execute actions without application-specific bindings. Uses COM Application Receivers to abstract Windows API complexity.

vs alternatives: More flexible than traditional RPA tools (UiPath, Automation Anywhere) because it uses LLM reasoning over visual state rather than rigid recorded macros, and more accessible than Selenium/Playwright because it works with any Windows GUI without requiring element selectors.

UFO³ Galaxy enables a Constellation Agent to decompose high-level tasks into subtasks, distribute them across multiple registered Windows devices, and coordinate execution through an Agent Interaction Protocol (AIP). The system maintains device lifecycle state (registration, heartbeat, availability), routes tasks to appropriate devices based on capability matching, and aggregates results. Task Constellation manages task dependencies and execution order across heterogeneous devices in a network.

Unique: Implements a two-tier agent hierarchy where Constellation Agent (Galaxy layer) performs task decomposition and device routing, while UFO² agents (device layer) execute concrete actions. Uses Agent Interaction Protocol (AIP) as a standardized communication layer between tiers, enabling loose coupling and independent scaling.

vs alternatives: Differs from monolithic RPA platforms (UiPath Orchestrator) by using LLM-driven task decomposition instead of pre-built workflows, and from simple multi-machine scripts by providing structured device lifecycle management and cross-device result aggregation.

UFO³ provides a web-based interface for submitting automation tasks, monitoring execution progress, viewing device status, and managing device registrations. The Web UI communicates with the Galaxy orchestrator via REST APIs, displays real-time execution logs and screenshots, and allows users to pause/resume/cancel tasks. Supports role-based access control for multi-user environments.

Unique: Provides a unified web interface for both task submission and device management, allowing users to view device status, capabilities, and execution logs in a single dashboard. Supports real-time updates via polling or WebSocket.

vs alternatives: More user-friendly than command-line interfaces because it provides visual feedback and forms. More integrated than separate monitoring tools because it combines task submission, execution monitoring, and device management.

UFO³ uses a hierarchical configuration system (YAML/JSON files) to define agent behavior, device capabilities, LLM provider settings, and knowledge base sources. Configuration files are organized by scope: agent-level (model selection, prompt templates), device-level (capabilities, resource constraints), and system-level (Galaxy settings, database connections). The system supports configuration inheritance and environment variable substitution, enabling flexible deployment across development, staging, and production environments.

Unique: Implements a hierarchical configuration system with agent-level, device-level, and system-level scopes, allowing fine-grained control over behavior. Supports configuration inheritance and environment variable substitution for flexible deployment.

vs alternatives: More flexible than hardcoded settings because configuration can be changed without recompilation. More organized than flat configuration files because it uses hierarchical scopes.

UFO² includes a User Interaction Module that pauses automation and requests human input when the agent encounters ambiguous situations or needs confirmation. The module can display screenshots with annotations, ask multiple-choice questions, or request free-form text input. Responses are injected back into the agent's context, allowing it to continue with human guidance. Supports both synchronous (blocking) and asynchronous (non-blocking) interaction patterns.

Unique: Integrates human interaction as a first-class capability in the automation pipeline, allowing agents to pause and request input without external orchestration. Supports both synchronous and asynchronous interaction patterns.

vs alternatives: More integrated than external approval systems because it's built into the agent loop. More flexible than fixed approval workflows because agents can request different types of input based on context.

UFO³ logs all execution details (actions, observations, LLM responses, tool results) to structured logs that can be analyzed for debugging and improvement. The system captures LAM (Learning from Automation Metrics) data including action success rates, LLM reasoning quality, and tool call patterns. Logs include screenshots, action traces, and full context at each step, enabling post-mortem analysis of failures. Supports log export in multiple formats (JSON, CSV) and integration with external analytics platforms.

Unique: Captures comprehensive execution data including screenshots, action traces, and LLM reasoning, enabling detailed post-mortem analysis. Supports LAM data collection for continuous improvement and metrics tracking.

vs alternatives: More comprehensive than simple error logs because it includes screenshots and full context. More actionable than raw logs because it supports structured metrics and LAM data collection.

UFO² supports both LLM-generated actions (click, type, navigate) and deterministic automation actions (MCP tool calls, COM API invocations, PowerShell scripts). The system routes actions through an Automation Framework that dispatches to appropriate executors: GUI actions go to the screenshot-annotation-action loop, while tool calls invoke registered MCP servers or COM Application Receivers. This hybrid approach allows agents to use LLM reasoning for complex UI navigation while offloading structured tasks (data extraction, API calls) to deterministic tools.

Unique: Implements a unified action dispatch system that treats GUI actions and tool calls as first-class citizens in the same execution pipeline. Uses an Automation Framework abstraction layer that allows agents to reason about both modalities without distinguishing between them, reducing cognitive load on the LLM.

vs alternatives: More flexible than pure GUI automation (Selenium, Playwright) because it can invoke APIs and tools directly, and more practical than pure API automation because it can handle UI-only applications. Differs from workflow orchestration platforms (Zapier, Make) by supporting visual automation alongside tool integration.

UFO² builds prompts that include desktop screenshots, extracted text (via OCR), and semantic UI annotations (element labels, bounding boxes, hierarchy). The Prompt System constructs multi-modal inputs by combining these modalities with task context and memory, then sends them to LLMs that support vision (GPT-4V, Claude 3.5). The system maintains a Prompt Component library that allows customization of how screenshots, OCR, and annotations are formatted and prioritized based on agent strategy.

Unique: Implements a Prompt Component architecture that decouples screenshot capture, OCR, annotation, and formatting, allowing agents to customize which modalities are included and how they're prioritized. Supports both full-screenshot and region-of-interest (ROI) prompting to optimize token usage.

vs alternatives: More sophisticated than simple screenshot-to-LLM approaches because it adds semantic annotations and OCR, reducing ambiguity. More flexible than fixed prompt templates because components can be composed and reordered based on agent strategy.

browser-use/browser-use | DeepWiki Loading... Index your code with Devin DeepWiki DeepWiki browser-use/browser-use Index your code with Devin Edit Wiki Share Loading... Last indexed: 17 May 2026 ( 933e28 ) Overview System Architecture Installation and Setup Quick Start Examples Agent System Agent Core and Execution Loop Message Manager and Prompt Construction Agent State and History Management System Prompts and Output Formats Skills Integration Agent Configuration and Settings Loop Detection and Behavioral Nudges Message Compaction System Memory and Follow-up Tasks Judge System and Trace Evaluation Browser Session Management BrowserSession Lifecycle Browser Profile Configuration SessionManager and CDP Session Pool Target and Frame Management Navigation and Tab Control Event-Driven Architecture Event System Overview Event Types Reference Watchdog Pattern and Base Classes Core Watchdog Implementations DOM Processing Engine DOM Tree Construction DOM Serialization Pipeline Interactive Element Detection Visibility Calculation and Coordinate Transformation Screenshot Highlighting System Browser State Summary Markdown Extraction and HTML Serialization Tools and Action System Tools Registry and Action Models Built-in Actions Reference Action Execution Pipeline Custom Tools and Extensions Click Action Deep Dive Input Action and Autocomplete Detection FileSystem Integration Br

System Architecture | browser-use/browser-use | DeepWiki Loading... Index your code with Devin DeepWiki DeepWiki browser-use/browser-use Index your code with Devin Edit Wiki Share Loading... Last indexed: 17 May 2026 ( 933e28 ) Overview System Architecture Installation and Setup Quick Start Examples Agent System Agent Core and Execution Loop Message Manager and Prompt Construction Agent State and History Management System Prompts and Output Formats Skills Integration Agent Configuration and Settings Loop Detection and Behavioral Nudges Message Compaction System Memory and Follow-up Tasks Judge System and Trace Evaluation Browser Session Management BrowserSession Lifecycle Browser Profile Configuration SessionManager and CDP Session Pool Target and Frame Management Navigation and Tab Control Event-Driven Architecture Event System Overview Event Types Reference Watchdog Pattern and Base Classes Core Watchdog Implementations DOM Processing Engine DOM Tree Construction DOM Serialization Pipeline Interactive Element Detection Visibility Calculation and Coordinate Transformation Screenshot Highlighting System Browser State Summary Markdown Extraction and HTML Serialization Tools and Action System Tools Registry and Action Models Built-in Actions Reference Action Execution Pipeline Custom Tools and Extensions Click Action Deep Dive Input Action and Autocomplete Detection FileS

Agent System | browser-use/browser-use | DeepWiki Loading... Index your code with Devin DeepWiki DeepWiki browser-use/browser-use Index your code with Devin Edit Wiki Share Loading... Last indexed: 17 May 2026 ( 933e28 ) Overview System Architecture Installation and Setup Quick Start Examples Agent System Agent Core and Execution Loop Message Manager and Prompt Construction Agent State and History Management System Prompts and Output Formats Skills Integration Agent Configuration and Settings Loop Detection and Behavioral Nudges Message Compaction System Memory and Follow-up Tasks Judge System and Trace Evaluation Browser Session Management BrowserSession Lifecycle Browser Profile Configuration SessionManager and CDP Session Pool Target and Frame Management Navigation and Tab Control Event-Driven Architecture Event System Overview Event Types Reference Watchdog Pattern and Base Classes Core Watchdog Implementations DOM Processing Engine DOM Tree Construction DOM Serialization Pipeline Interactive Element Detection Visibility Calculation and Coordinate Transformation Screenshot Highlighting System Browser State Summary Markdown Extraction and HTML Serialization Tools and Action System Tools Registry and Action Models Built-in Actions Reference Action Execution Pipeline Custom Tools and Extensions Click Action Deep Dive Input Action and Autocomplete Detection FileSystem I

browser-use/browser-use | DeepWiki Loading... Index your code with Devin DeepWiki DeepWiki browser-use/browser-use Index your code with Devin Edit Wiki Share Loading... Last indexed: 17 May 2026 ( 933e28 ) Overview System Architecture Installation and Setup Quick Start Examples Agent System Agent Core and Execution Loop Message Manager and Prompt Construction Agent State and History Management System Prompts and Output Formats Skills Integration Agent Configuration and Settings Loop Detection and Behavioral Nudges Message Compaction System Memory and Follow-up Tasks Judge System and Trace Evaluation Browser Session Management BrowserSession Lifecycle Browser Profile Configuration SessionManager and CDP Session Pool Target and Frame Management Navigation and Tab Control Event-Driven Architecture Event System Overview Event Types Reference Watchdog Pattern and Base Classes Core Watchdog Implementations DOM Processing Engine DOM Tree Construction DOM Serialization Pipeline Interactive Element Detection Visibility Calculation and Coordinate Transformation Screenshot Highlighting System Browser Sta