GPTBots WorkSpace — Product Documentation

In one sentence: GPTBots is an enterprise AI agent and workflow platform. WorkSpace is its execution module — an out-of-the-box AI workspace for a company's internal employees, taking AI from "able to answer" to "able to act" and turning AI into real business outcomes.

Our Belief: From AI Assistance to AI Action

AI should not merely answer questions — it should deliver real business outcomes. WorkSpace covers the full range of enterprise AI execution scenarios through four native capabilities:

Module	Positioning	In One Sentence
Work	AI clone / autonomous execution	An Agent system with an autonomous loop that understands tasks, calls tools, and collaborates across devices
Search	Knowledge search / knowledge retrieval	Every answer has context: unified retrieval across documents, email, chat, code, and databases
Workflow	Intelligent workflow / process automation	Orchestrate in the development space, invoke with one click in the workspace, and speed up internal enterprise processes instantly
Agents	Multi-agent / orchestrated collaboration	Internal build-and-publish + third-party A2A integration, with shared context and memory

What Can I Do With It

WorkSpace is not just a chat tool — it proactively calls tools to complete tasks instead of only outputting text suggestions:

What You Want to Do	How WorkSpace Helps You
Organize a work directory stuffed with files	List files → classify by content → rename → move into the corresponding folders
Consolidate 50 invoices into an Excel sheet	Read each PDF → extract key fields → generate a structured table → export
Write a script for next week's launch event	Search industry materials → draft an outline → iterate repeatedly → output Word/PPT
Automatically monitor a competitor's website every day at 9:00	Create a scheduled task → crawl the page → compare changes → send to DingTalk/Feishu
Collaborate with AI directly in a Telegram group	Connect the channel → @Bot in the group → all group members can use the same Agent

Key difference: A traditional chatbot only answers questions; WorkSpace completes tasks. It features an autonomous reasoning-execution loop, 19 built-in tools, 21 preset skills, three-dimensional memory, and multi-device collaboration, and it can connect to 14+ IM platforms.

Three Usage Forms

Form	Best For	Core Capabilities
Web (browser)	Everyday conversations, workspace management, remotely initiating tasks	Executes by connecting to registered APP nodes through the Gateway
APP (macOS/Windows desktop)	Local file operations, deep development, IM channel integration, scheduled tasks	Runs the full Agent engine locally with full file system permissions
IM channels (Telegram/DingTalk/Teams, etc.)	Invoking AI within familiar communication tools	The Agent responds directly in groups or private chats

All three forms share the same workspace, the same set of memories, and the same execution engine.

🚀 First Time Here (Recommended Starting Point)

Document	Best For
Product Overview and Core Capabilities	Understand the product positioning, core capabilities, and target users
Quick Start	Run your first task in 5 minutes (covers Web, APP, and IM entry points)
Web vs. APP Capability Comparison	Choose the usage form that fits you

💬 WorkSpace Pages (Everyday User Use)

Organized by the workspace sidebar menu, covering both the Web and APP:

Document	Menu	Description
Work Conversation Overview	Work	Core conversation interface, welcome page, quick actions
Conversation Interaction	Work	Message stream, streaming output, session management
Multimodal Input and Output	Work	Image/video/document input, 9 file preview types
Sub-agents	Work	Task decomposition, sub-agent dispatch, cross-node execution
AI Search	Search	Intelligent search (LLM-based open-ended search)
Agents	Agents	A2A protocol agents + traditional Agent management
Workflow	Workflow	Visual workflow orchestration and execution

🛠️ APP Settings (Desktop APP Only)

Settings menus exclusive to the desktop APP. Web users cannot see these; the corresponding enterprise-level configuration lives in "WorkSpace Management":

Document	Menu	Description
APP Settings Overview	Settings entry	Menu navigation and overall layout
APP Quick Start	—	APP installation, login, and first conversation
Model Configuration	Models	Multiple providers, API Keys, capability tags
Skill Management	Skills	21 presets + custom + enterprise skills
Tool Management	Tools	19 built-in + MCP + API tools
Sub-agent Management	Sub-agents	Pre-registered sub-agent configuration
Channel Configuration	Channels	14+ IM platform integrations
Scheduled Tasks	Scheduled Tasks	NEW One-time / recurring / Cron automation
Memory Management	Memory	Account-level memory and graph visualization
Runtime Security	Runtime Security	Tool permissions, nodes, sandbox

🏢 WorkSpace Management (Administrators)

The enterprise administrator's console — centralized management of resources, permissions, security, and branding:

Document	Menu	Description
Workspace Management Overview	Management entry	Navigation structure and permission system
Usage Statistics	Usage	Overview, Work, and AI Search usage
Work Node Management	Work	Real-time node list, types, and filtering
Enterprise Resources	Enterprise Resources	Skills, tools, A2A, Agents, Workflow
Log Audit	Log Audit	Work, Search, and Agents operation logs
Security Guardrails	Security Guardrails	Tool permissions, content moderation, data anonymization
Member Permissions	Member Permissions	Member management, roles, fine-grained authorization
Advanced Settings	Advanced Settings	Branding, themes, enterprise memory, integration entry

⚙️ Agent Engine (Deep Dive Into the Underlying Layer)

Cross-page underlying technical capabilities. Regular users do not need to read this; consult it only if you are interested in the architecture or troubleshooting:

Document	Description
Agent Loop Engine	25-round loop, circuit breaker, streaming events, snapshot recovery
Context 5-Layer Compression	Tool truncation, message pruning, LLM summarization, degradation notices
Three-Dimensional Memory System	Account/enterprise/session dimensions, knowledge graph, memory injection
Multi-Node Architecture	Node discovery, Gateway routing, cross-account isolation

📚 Appendix

Document	Description
Supported Model List	Complete models + context windows + capability matrix
Channel List and Capability Matrix	14+ channels + multimodal support comparison
Glossary	Definitions of key technical terms

🛒 AI Marketplace

A range of AI applications officially developed by GPTBots, covering scenarios such as meetings and office work:

Document	Description
Marketplace Overview	AI Marketplace overview and management
Live Speechly	Real-time multilingual meeting translation AI application (macOS / Windows)
WorkPilot	A plugin that embeds Agent capabilities into Microsoft Word

📱 Mobile Integration (Enterprise APP Embedding)

Integrate the entire WorkSpace into an enterprise's mobile APP for login-free employee access:

Document	Description
Mobile Integration Overview	Architecture, URL rules, WebViewBridge event protocol
AiToken Encryption	Generate AiToken by encrypting the email with an RSA public key (includes Java / TS code)
Android Integration	Permissions, WebView, JSBridge, file selection
iOS Integration	WKWebView, permissions, file selection, message bridging

Reading Paths for Different Roles

I Am a Regular User (First Encounter)

Product Overview
Quick Start
Work Conversation Overview
(If you want to use the desktop APP) APP Quick Start

I Am an Advanced APP User (Want Deep Automation)

I Am a WorkSpace Administrator

Workspace Management Overview
Member Permissions
Security Guardrails
Enterprise Resources
Advanced Settings (brand customization, enterprise memory, integration entry)

I Am a Technical Architecture Enthusiast

About This Documentation

This documentation is written based on the actual code of the Web (React + Vite) and the desktop APP (Tauri 2 + Node.js Sidecar)
All technical parameters (round limits, timeouts, thresholds, etc.) are derived from code constants
Content marked NEW represents capabilities added in recent versions
Screenshots in this documentation are continuously updated; explanatory text labeled screenshot location marks image placeholders that are to be added