YggLab AI Agent Desktop
One desktop where 20+ CLI agents actually collaborate — assign work solo, form a team, or command them from your phone while work keeps moving 24/7.
In short
YggLab is an AI agent collaboration desktop that runs on your own machine. It auto-detects the 20+ command-line agents already installed locally — Claude Code, Codex, Gemini CLI and others — and lets them work in parallel or as a team, alongside 20+ built-in assistants for slides, spreadsheets, writing and design that work out of the box. Conversations and task logs stay on the device, and models are called directly through their official APIs with no relay server in between.
What the agent takes off people's plates
Core capabilities
20+ CLI agents in concert
Detects the 20+ command-line tools already on your machine — Claude Code, Codex, Gemini CLI and more — and reuses them as-is: no reinstall, no overwriting your existing config. Run them in parallel with separate remits, or assemble a squad (strategy + copy + media buying) that drives one whole task to completion in a single window.
20+ built-in assistants, zero setup
Useful even with no CLI installed at all: slide decks, Morph-animated presentations, investor pitch decks, Excel analysis, dashboards, financial models, Word documents, academic writing, Mermaid diagrams, UI/UX design, project planning, autonomous Cowork execution — 20+ specialists, each with its own remit. You can also spin up your own assistant tuned to how your team works.
Away from the desk, still at the console
Send instructions over Telegram, WeChat, Feishu, DingTalk or the built-in WebUI. You are on the move while the agents keep working on your own machine — deliverables still land locally, ready for you to pick up and refine.
Set it once, it runs around the clock
Schedule agents like cron jobs: a code review every morning, a project backup every night, a sales report or stand-up note every Monday. The task list shows each job’s status and the agent handling it.
Any vendor, any model, no lock-in
Any vendor, any endpoint, any API key — 30+ platforms including OpenAI (GPT), Anthropic (Claude), Google (Gemini), DeepSeek, Qwen and local Ollama models. Mix them on one desktop: a different model per agent, switchable at any time. Add an MCP server once and it syncs to every agent; the Skills marketplace extends what the assistants can do.
Privacy first — data never leaves your machine
Conversations and task logs live in a local SQLite database with no cloud sync; API keys are encrypted in the system keychain; models are reached directly through their official APIs with no proxy or relay; PDF and Word extraction and generation run entirely on-device. Native .pptx / .xlsx / .docx / .pdf output lands straight on your desktop.
Side by side
YggLab vs. editor-bound and single-model AI tools (as stated in YggLab’s official FAQ)
| Dimension | YggLab | Cursor / Copilot | Claude Desktop |
|---|---|---|---|
| Who it serves | Everyday work across the whole desktop — slides, spreadsheets, writing, design, code — usable without being a developer | Primarily helps developers write code, centred on the editor | A general chat assistant, but confined to its own client |
| Model choice | Any vendor, endpoint and API key — 30+ platforms, mixable, a different model per agent | Limited to each vendor’s own model line-up | Claude only — a single model |
| Multi-agent collaboration | 20+ CLI agents run in parallel or as a squad — Claude, ChatGPT and Gemini working together at once | A single assistant working around the current codebase | A single assistant session, no cross-model teaming |
| Once you leave the machine | Assign work via Telegram / WeChat / Feishu / DingTalk / WebUI; scheduled jobs keep running 24/7 | Built around interactive use at the editor | Built around an open client session |
| Where data lives | Conversations, files and task logs all in local SQLite; API keys in the system keychain; direct model calls with no relay | Depends on each vendor’s cloud policy — check their privacy terms yourself | Depends on the vendor’s cloud policy — check their privacy terms yourself |
Subjects and capability matrix
Covered modules
Agent collaboration
20+ CLIs · parallel or team
Built-in assistants
20+ specialists · no setup
Remote control
Telegram / WeChat / Feishu / DingTalk / WebUI
Scheduled tasks
Unattended, 24/7
Model configuration
30+ platforms · per-agent models
Skills marketplace
Install community skills · build your own
Unified MCP
Configure once · all agents connected
Office toolkit
Native .pptx / .xlsx / .docx / .pdf
How the agent and human supervision collaborate
System architecture (conceptual)
human
You (desktop or phone)
source
Telegram / WeChat / Feishu / DingTalk / WebUI
source
Local files and working directory
agent
YggLab local engine (scheduling / cron / MCP)
agent
Installed CLI agents (Claude Code / Codex / Gemini CLI …)
agent
20+ built-in assistants
source
Model APIs (your own keys, direct, no relay)
output
Local SQLite (conversations and task logs)
output
Deliverables on your desktop (.pptx / .xlsx / .docx / .pdf)
Flow
- You (desktop or phone)→Telegram / WeChat / Feishu / DingTalk / WebUIassign remotely
- Telegram / WeChat / Feishu / DingTalk / WebUI→YggLab local engine (scheduling / cron / MCP)
- You (desktop or phone)→YggLab local engine (scheduling / cron / MCP)chat locally
- Local files and working directory→YggLab local engine (scheduling / cron / MCP)
- YggLab local engine (scheduling / cron / MCP)→Installed CLI agents (Claude Code / Codex / Gemini CLI …)dispatch / team up
- YggLab local engine (scheduling / cron / MCP)→20+ built-in assistants
- Installed CLI agents (Claude Code / Codex / Gemini CLI …)→Model APIs (your own keys, direct, no relay)direct call
- 20+ built-in assistants→Model APIs (your own keys, direct, no relay)
- YggLab local engine (scheduling / cron / MCP)→Local SQLite (conversations and task logs)local only
- Installed CLI agents (Claude Code / Codex / Gemini CLI …)→Deliverables on your desktop (.pptx / .xlsx / .docx / .pdf)
- 20+ built-in assistants→Deliverables on your desktop (.pptx / .xlsx / .docx / .pdf)
Verifiable facts
Numbers and sources
The current public version is YggLab v0.1.29.
Auto-detects and hooks into 20+ locally installed CLI agents (Claude Code, Codex, Gemini CLI and others), reusing existing configuration without reinstalling or overwriting it.
Supports 30+ model platforms — OpenAI (GPT), Anthropic (Claude), Google (Gemini), DeepSeek, Qwen and local Ollama models — mixable per agent on the same desktop.
Ships with 20+ out-of-the-box assistants spanning slides, Excel, Word, academic writing, diagrams, UI/UX design and project planning — usable without any coding background.
100% local execution: conversations and task logs stay in a local SQLite database with no cloud sync, API keys sit in the system keychain, models are called directly with no relay, and PDF / Word extraction and generation run entirely on-device.
System requirements are Windows 10+, macOS 12+ or Ubuntu 18.04+; an ordinary office machine suffices, since the app itself is lightweight and inference happens in cloud APIs or a local model.
Work can be assigned remotely via Telegram, WeChat, Feishu, DingTalk or the built-in WebUI, with scheduled jobs running unattended around the clock.
Delivery and pricing
Pricing
Contact sales for a quote · available for macOS, Windows and Linux
YggLab does not publish list prices. Its official FAQ states plainly that you should contact a sales representative for a quote — reach out to the YGG sales team with your team size and use case, and we will come back with pricing and how to obtain the installer.
Provided by the client
- · Model inference: bring your own vendor API keys or a self-hosted Ollama model; token spend is on the customer
- · Subscriptions or licences for the CLI agents themselves (Claude Code, Codex, Gemini CLI account fees and the like)
Scan for the full solution document
Scan with WeChat to reach the solution lead and request the complete documentation and deployment checklist.
Where it fits
When this approach does not apply
- Reusing Claude Code / Codex / Gemini CLI assumes those CLIs are already installed locally — YggLab discovers and reuses them, it does not install them for you. With none installed you can still use the 20+ built-in assistants, but the parallel and team multi-agent value largely disappears.
- Model inference is not included. YggLab calls official APIs directly, so you need your own keys from OpenAI / Anthropic / Google / DeepSeek / Qwen and the like, or a self-hosted Ollama model; token spend is on you.
- The agents run on your own machine: when it is shut down, asleep or offline, neither remote commands nor scheduled jobs execute. Teams with no always-on machine — fully mobile setups — are not a fit.
- The stated system requirements are Windows 10+, macOS 12+ or Ubuntu 18.04+. Older systems, browser-only environments and locked-down endpoints that forbid installing desktop apps are out of scope.
- Every capability and figure here comes from YggLab v0.1.29’s own website; none has been independently benchmarked or security-audited. Run a pilot against your own workflow before committing.