ColineColineColineDocs

Kairo

Kairo is the AI assistant built into Coline. It can read your workspace context (tasks, notes, calendar events, boards, emails, messages, and files) and take actions on your behalf. Use it for planning, writing, searching, creating, and automating work across every Coline app.

Chat interface

The main Kairo experience is a full-page chat at Chat in the sidebar. The interface has three areas:

  • Conversation sidebar (left): Lists your past conversations with timestamps. Click any conversation to resume it, or start a new one.
  • Message area (center): Shows the back-and-forth between you and Kairo. Messages stream in token-by-token so you can read responses as they generate.
  • Input bar (bottom): Where you type prompts. Supports file attachments, @mentions of workspace items, and toggles for model settings.

Models

Kairo gives you access to multiple AI models. Each model has different strengths. You can switch models at any time using the model selector in the input bar, or choose Auto to let Kairo pick the best model for your prompt.

ModelContextWeb searchImage genToolsReasoningUsage
GPT-5.2400KYesYesYesInstant / Low / Medium / High1x
GPT-5.1400KYesYesYesInstant / Low / Medium / High1x
GPT-5400KYesYesYesInstant / Low / Medium / High1x
GPT-5 Mini128KYesYesYesInstant / Low / Medium / High0.5x
GPT-5 Nano64KYesYesYesInstant / Low / Medium / High0.5x
Claude Opus 4.61MNoNoYesToggle + Low / Medium / High / Max3x (6x on Max)
Claude Opus 4.5200KNoNoYesToggle3x
Claude Sonnet 4.5200KNoNoYesToggle1.5x
Kimi K2.5256KNoNoYesToggle1x
Kimi K2 Thinking256KNoNoYesAlways on1x
DeepSeek V3.2128KNoNoYesAlways on1x
DeepSeek V3.2 Speciale128KNoNoYesAlways on1x
Grok 4 Fast2MNoNoYesToggle1x

The Usage column shows how each model affects your quota. A 0.5x model uses half the quota per request, while a 3x model uses three times the quota. Claude Opus 4.6 uses 6x when thinking effort is set to Max. Pro plans include 1,500 credits and Team plans include 1,700 credits.

Reasoning

Some models support adjustable reasoning, which lets the model spend more time thinking through complex problems before responding.

  • GPT models: A slider with four levels controls reasoning depth: Instant (no extra reasoning), Low, Medium, and High (deepest analysis).
  • Claude Opus 4.6: Toggle thinking on/off, then choose adaptive effort: Low, Medium, High, or Max.
  • Claude Sonnet 4.5 / Opus 4.5: Toggle thinking on/off. When enabled, Kairo uses a fixed 32k thinking budget.
  • Kimi models: K2 Thinking always have reasoning enabled, while K2.5 has is toggleable.
  • DeepSeek models: Both V3.2 and V3.2 Speciale always have reasoning enabled.
  • Grok: A toggle switches between reasoning and non-reasoning variants.

When web search is enabled, Kairo can query the web for up-to-date information before responding. Search results appear as grounded citations in the response. Toggle web search on or off from the settings in the input bar.

Available on OpenAI models only.

Image generation

When image generation is enabled, Kairo can create images from text descriptions. Generated images appear inline in the conversation. Toggle image generation on or off from the settings in the input bar.

Available on OpenAI models only.

Workspace tools

Kairo can read and write across your entire Coline workspace. When you ask Kairo to do something that requires accessing your data, it uses tools to complete the action. Tool calls appear as inline indicators in the response so you can see what Kairo accessed.

Available tools include:

  • Boards: Find boards, create and update cards, manage columns and labels, add comments and checklists
  • Calendar: List calendars, find and create events, update existing events
  • Notes: Search notes, view details, create and update notes, append content
  • Issues: Find issues, view details, create new issues
  • Mail: Search email, read threads, draft and send emails, categorize and batch-manage messages
  • Messages: List rooms, send messages
  • Files: Search files, view details, create folders, move and delete files
  • Code: Read, write, and edit files in the code editor, run terminal commands, search codebases
  • Images: Generate and edit images

Tool calls are classified for safe execution. Read-only operations like searching and listing can run in parallel. Mutations like creating or updating items run sequentially to keep things predictable.

More tools and integrations will be added over time.

@mentions

Type @ in the input bar to mention a specific workspace item. Kairo pulls in that item's content as context for your conversation. You can mention tasks, notes, events, boards, emails, and other items. The mention appears as a chip in your message, and Kairo receives the full details of the referenced item.

File attachments

Drag files into the input bar or click the attachment button to upload files. Kairo accepts up to 10 files at once, with a maximum of 50 MB per file. Supported types include PDFs (text is extracted), images (analyzed visually), and common document formats.

Multi-model comparison

Compare responses from different models side by side. When comparison mode is active, Kairo sends your prompt to multiple models simultaneously and displays their responses in parallel columns. This is useful for evaluating which model handles your specific task best.

Quick Kairo

Press Cmd + Shift + O (Mac) or Ctrl + Shift + O (Windows/Linux) to open Quick Kairo, a floating chat panel that appears over any screen in Coline. It supports the same features as the full chat (model selection, file attachments, mentions, web search, and image generation) without navigating away from your current work.

Quick Kairo is also available from the Ask Kairo button on the Home dashboard.

Conversations

Kairo automatically saves your conversations. Open the conversation sidebar to browse, resume, or delete past chats. For long conversations, Kairo compacts older messages to stay within the model's context window while preserving the key points.

Usage

Kairo usage is tracked per workspace based on your plan tier. Each model has a usage multiplier that affects how quickly your quota is consumed. You can monitor your usage in workspace settings.

Tips

  • Be specific with context. Use @mentions to point Kairo at the exact items you need help with rather than describing them.
  • Match the model to the task. Use GPT-5 Nano or Mini for quick questions and drafts. Use Claude Opus or GPT-5.2 for complex reasoning and analysis.
  • Use Quick Kairo for fast lookups. The keyboard shortcut opens a chat panel without leaving your current workflow.
  • Enable web search for current information. Kairo's training data has a cutoff, so toggle web search on when you need recent facts, news, or documentation.
  • Check tool indicators. The inline tool badges show what Kairo accessed, so you can verify it pulled the right context.