YggLab features

What can the YggLab chat window do?

The chat window has 15 capabilities, including side questions, queued instructions, three permission tiers and full visibility into every tool call. It isn't a box you type into and then wait — while a task runs in the background you can interject, queue up what's next, dial back permissions, and see what it's doing at each step.

• ★Ask on the side  Ask a quick question about the current context while the main task keeps running. The answer comes back separately and the main task is unaffected.
• ★Queue instructions  Queue up the next few steps while the AI is still answering. Queued items can be edited, reordered or removed.
• ★Permission tiers  Three tiers — plan only, auto-accept edits, fully autonomous — chosen by how much you trust it. Any single tool call can be allowed once or from then on.
• ★Every tool call is visible  What command ran, in which directory, the exit code, how long it took, which files changed and how many lines — every call laid out as a card.
• Switch engines mid-chat  Switch to a different CLI mid-conversation. The context is compressed into a handover summary and carried over; the original conversation stays intact.
• Automatic failover on quota  When a model runs out of quota or its API errors, you're told and it switches to one that works instead of stalling.
• Voice input  Talk instead of typing, or drop in an audio file to transcribe.
• @ for files · / for commands  Type @ to pull a project file into the conversation, / to open the command menu.
• Full-text history search with jump-to  Search every past message by keyword and jump straight to it.
• Conversation minimap  An index of turns beside long conversations, so you can find the one you mean.
• Live context usage  See how much context this turn is using before it runs out.
• Export a conversation  Copy the whole thing out or save it to a file.
• Bulk chat management  Select several and export them as a ZIP (transcripts plus project files), or delete in bulk.
• Pin · rename · group by date  Pin the ones you use; the rest file themselves under Today, Yesterday and Last 7 days.
• Organize chats by project  Every conversation about the same piece of work lives under one project.

How do I edit what the AI produced?

The output panel isn't just a preview. Slides, spreadsheets, documents, images, HTML and PDFs can all be edited in place, diffed and rolled back to an earlier version without switching to another application.

• ★Edit in the preview  Whatever the AI got wrong, fix it in the preview panel and save — no switching apps.
• ★Change panel  See exactly which files were touched and how many lines added or removed, and discard changes one by one.
• ★Snapshots and history  Roll back to an earlier version if something breaks, and snapshot manually at important points.
• Live refresh as it writes  Watch a document take shape as it's written; newly generated office files open in the preview on their own.
• Live HTML preview with element picker  See a page as it's built, and click any element in the preview to jump to the code behind it.
• PDF and Mermaid rendering  PDFs just open, and flowcharts and sequence diagrams render as diagrams instead of a wall of code.
• Tabs · split view · open in a system app  Compare several files, split the view, or hand one straight to Word or Keynote.

How far do scheduled tasks go?

Scheduled tasks can push results to your chat app, run in either of two context modes, keep a history with failure reasons, be triggered by hand, and stop the computer sleeping. Each task can also have its own project folder and model.

• ★Results pushed to your chat app  The finished report goes to whichever recipient you named on Slack, Telegram or another channel, so you see it even when you're away.
• ★Two execution modes  Routine checks start a fresh conversation each run; long-running work keeps accumulating context and picks up where it left off.
• Run history and failure reasons  Whether each run succeeded, when the next one is due, and why a run failed.
• Run now  Trigger a run by hand instead of waiting for the schedule.
• Keep the computer awake  Block sleep in one click so a task isn't skipped because the machine dozed off.
• Per-task folder and model  Where a task works and which model it uses can differ from every other task.

Which remote channels are supported?

Seven chat channels are supported — Slack, Discord, Telegram, WeChat, WeCom, Feishu and DingTalk — plus a built-in WebUI you sign into by scanning a QR code. A message on any of them assigns work to the agents on your computer, and files can be sent and received in the chat.

• ★Seven chat channels  Slack, Discord, Telegram, WeChat, WeCom, Feishu and DingTalk are all supported.
• ★Send and receive files in chat  Drop in a PDF, Word file or image and it starts working; output comes back to the same chat, with large files split automatically.
• ★Sign into the WebUI by QR code  Scan a code with your phone — no username or password to type.
• A different agent and model per channel  One agent on Slack, another on Telegram.
• Pairing approval and access control  When someone messages the bot you have to approve them first, and access can be revoked at any time.
• Manage your own WebUI account  Change the username or password, or reset a forgotten one in a click.
• Temporary public share links  Generate a temporary externally reachable address; it stops working when the app restarts, and the interface warns you about the risk.
• Connect a remote agent  Bring an agent running on another machine or server into the same workspace.

How do several agents form a team?

Build a team by picking specialties — Documents, Design and visualization, Business, Development — and the engine matches an agent to each. Or start from a ready-made template: debate, deep interview, expert panel.

• ★Pick by specialty  Someone to write the docs, someone to do the visuals — pick by area and the engine matches an agent. Pick two who overlap and it tells you the extra member only adds coordination cost.
• ★Ready-made team templates  Debate, deep interview, expert panel — click one when you're not sure what a team is for.
• Team management  Pin the teams you use often; swap a member or its engine when it isn't working out.
• A leader assigns, members cross-check  The leader breaks the task down and hands out the pieces; members can question each other, pass files around and cross-check the results.
• Agents working in parallel  20+ CLI agents can work on one desktop at the same time, each on its own piece.

Which assistants are built in?

There are 20 built-in assistants across four specialties — Documents, Design and visualization, Business and Development — plus an office setup assistant that ships with the app. They work immediately, with no prompt writing, and teams are assembled from the same four groups.

How do I create my own assistant?

You can build an assistant by talking to it. Not sure how to write a prompt? It asks a few questions and drafts the rules for you. You can also turn the current conversation into an assistant, or copy a built-in one and adjust it.

• ★Build an assistant by talking  It asks you a few questions and drafts the rules for you; you can also turn the conversation you're in into an assistant.
• Copy a built-in one and adjust  Like one of the built-ins but want it slightly different? Duplicate it and make it yours.
• 20+ assistants, ready to use  Slides, spreadsheets, documents, academic writing, UI design, data dashboards, financial modeling, diagrams.

How do skills work?

A skill installed once is available to every assistant, with no per-assistant setup. You can also capture the process you just worked out as a new skill, or import one from a local folder or zip.

• ★Install once, every assistant has it  A skill installed once is available to every assistant — no per-assistant setup.
• ★Turn what you just did into a skill  Capture the process you just worked out as a skill or rule, and run it the same way next time.
• Import a local skill  Drag in a skill you downloaded elsewhere — a folder, a parent folder or a zip — or point at a skills folder on your machine.
• Suggested skills  Not sure which skills to enable? Let it pick a set based on what the assistant is for.
• Skill marketplace  Browse and install community skills, or build and share your own.

How do I set up models and accounts?

Either route works. Sign in to a YggLab account and cloud models are configured for you. Or bring your own API key: 28 model platforms are one click away, and you can point at any OpenAI- or Anthropic-compatible endpoint.

• ★Works without hunting for a key  Sign in to an account and cloud models are configured for you — nothing to register for.
• ★Token usage  Today's usage, the last 7 days, which model costs the most, and a line-by-line record.
• 28 built-in model platforms  OpenAI, Anthropic, Google Gemini, DeepSeek, Alibaba Qwen, Moonshot Kimi, Zhipu GLM and local models via Ollama — all one click.
• Top up online  Add credit and see your balance and every charge.
• Key rotation and bulk testing  Add several keys and it rotates between them; test them all in one click and clear the dead ones.
• Protocol auto-detection  Enter an endpoint and key and it works out the API format, telling you if you've picked the wrong platform.
• Engine health checks and failover  When the current engine misbehaves it measures latency and availability and moves you to the best one.

How do I connect MCP tools?

Add an MCP server once and every agent picks it up — no per-agent setup. MCP tools you configured in another CLI can be scanned and imported in one click.

• ★Import MCP from CLIs you already use  Tools you configured in another CLI can be scanned and brought over without reconfiguring.
• One MCP setup for every agent  Add an MCP server once and every agent picks it up.
• Manual runs and history  Run a connected tool by hand to try it, and the record is kept.
• MCP services that use OAuth  MCP servers requiring OAuth authorization are supported.
• Add a custom ACP agent  Another AI CLI on your machine can be added as an engine — simple mode or advanced JSON, with environment variables and a connection test.

What does the project workspace do?

The project workspace lets you drag files in from Finder or File Explorer, search the file tree, preview, rename and download from a context menu, and open the project in VS Code, a terminal or your file manager in one click.

• Drag and drop  Drag files straight in from Finder or File Explorer.
• File context menu  Preview, rename, download, add to the current conversation, reveal in folder.
• File-tree search and scratch space  Search the project file tree, plus a scratch area for things produced along the way.
• Open in an external tool  Open the project in VS Code, a terminal or your file manager in one click.

What's the desktop experience like?

YggLab is a proper desktop application: it lives in the system tray, starts at login, updates itself, speaks nine languages, has five colour schemes plus light/dark following the system, supports custom CSS themes, and comes with a desktop companion that can handle permission prompts.

• ★Lives in the system tray  Minimised to the tray you can still start a conversation, see what's running, and pause everything in a click.
• ★Launch at login · close to tray  So a scheduled task isn't missed just because you closed the window.
• ★In-app updates  Download and install new versions from inside the app, with pre-release builds optional.
• First-run setup wizard  Three steps on first launch, and a Claude Code install you're already signed into is detected automatically.
• A purpose-first home screen  It opens asking who should take this one — pick an assistant or an example and start, instead of staring at an empty box.
• A “working on it” panel  See what's currently running right from the home screen.
• Nine interface languages  English, Simplified and Traditional Chinese, Japanese, Korean, Turkish, Russian, Ukrainian, Portuguese.
• Five colour schemes, light and dark following the system  Pick a different palette, or let the interface follow your system between light and dark.
• Custom CSS themes  Write your own styles, add a background image, save it as a theme and switch at will.
• Zoom and text size  Chat text, Markdown and code blocks each have their own size setting.
• Desktop companion  A little companion on your desktop that can handle permission prompts, with a do-not-disturb mode.
• Custom working and log folders, plus cache clearing  You decide where files live, and the cache clears in one click.

How does YggLab protect your privacy?

YggLab runs entirely on your own computer. Conversations, project files and output never pass through YGG Lab's servers; API keys are stored locally and used to reach model vendors directly. Which model you use and where your data goes is entirely your decision.

• Everything runs locally  The app runs on your own computer; conversations and files are not uploaded.
• Keys talk to vendors directly  Nothing is relayed through us, and there is no forwarding proxy.
• Office files generated natively  Slides, spreadsheets, documents and PDFs are produced locally, with no third-party online service.
• Share links stay under control  A temporary share address expires when the app restarts, and the interface warns you about the risk.