Working with AI

There are several ways to start a conversation with AI in Tana:

• Today: The "Ask or Capture…" bar below the date header in Today. Each message starts a fresh solo chat pinned to the selected date.
• Space home: The "Ask or start a chat..." bar at the bottom of every space home. Type a message and press enter to create a new chat scoped to that space.
• Meetings: Every meeting has two chat surfaces. The private Ask or Capture pill in the bottom bar is yours alone and the AI auto-responds to every message. The shared Group Chat, opened from the call toolbar, is visible to all attendees and stays quiet until you mention @Tana.
• Quick chat: Press Cmd+LCtrl+L on the desktop app to open a new chat from anywhere. Press Cmd+Shift+LCtrl+Shift+L to open it with voice dictation active.
• Sidebar: Click "New Chat" in the sidebar.
• Create menu: Select "Chat" from the create menu.
• Documents: Many document types have a contextual chat panel that scopes the conversation to the open document.

Recent chats appear in the sidebar sorted by last activity, with unread indicators for new messages.

In your workspace

The AI has full access to your workspace content (within your permissions). It can:

• Read: Documents, types, fields, meeting transcripts and participants, screen capture screenshots, calendar events, skill definitions, previous chats, tables, attached PDFs, and uploaded audio and video
• Search: Keyword search, semantic search, relationships between documents
• Create: Docs, tasks, typed entries, types with fields, skills, artifacts, and AI-generated images
• Update: Titles, content, fields, calendar events, space configuration, table data, and type definitions
• Delete: Any document

Updates and deletes and creates all appear as proposals for your review. AI never changes your workspace without your approval.

Connected tools

When you connect integrations, the AI can act on them from chat:

• GitHub: Search and create issues, review and merge pull requests, read code and files, comment on issues and PRs
• Slack: Send messages to channels or people, list channels and users
• Linear: Search, create, and update issues
• Coding tools: Hand off the current document or chat as a self-contained prompt to Claude Code, Cursor, Codex, GitHub Copilot, Lovable, or v0. Triggered by phrasings like "create a PR" or "send this to Cursor"; when more than one tool is enabled, you pick.

Configure integrations in Settings. See Integrations.

Managing access

The AI can also propose changes to a document's access: restricting it to specific participants with editor or admin roles, opening it back up to everyone in its space or event, and toggling the Guest Link (anyone with the link can view). Access changes go through the same proposal approval flow as the rest of the AI's edits, and the proposal preview shows a clear access callout before you accept.

Default content language

Every organization picks a default content language in Settings, Organization, Language. The AI writes its replies, generated artifacts, outcomes, summaries, and document titles in that language no matter which language the conversation is happening in. Meeting transcription stays multilingual (auto-detected per speaker) so a mixed-language call still transcribes everyone correctly.

Usage limits

Each organization has monthly and rolling short-window AI usage limits, plus a shared org-wide pool. As a bucket approaches 80 percent, an amber "Approaching usage limit" callout appears at the top of every chat panel and a soft warning lands in Activity. When a bucket is exceeded, the AI returns a red "Usage limit reached" callout instead of a reply. Your own monthly and 5-hour caps are visible as progress bars in Settings, User, Usage.

Web search and image generation

Models that support web search can look up current information when your question goes beyond your workspace. Results include citation links to sources. Web search availability varies by model.

The AI can also generate images from text descriptions. Generated images become image documents in your workspace.

Pending proposals

When AI generates changes, proposals appear as inline cards on the message that produced them, so you can see what the AI did and decide on it in place. While AI is still thinking, a "Refining proposals" indicator shows progress. Proposals are grouped on the card by operation (deletes first, then updates, then creates) and by document type within each group.

Accepting and discarding

For each proposal:

• Accept applies the change to your workspace
• Discard rejects the proposal without applying it

Click any proposal to open the detail view with the full content of the proposed change. Proposals are editable while pending, so you can modify content directly before accepting.

For updates, three view modes are available:

• Diff: Side-by-side comparison of original and proposed (default)
• Proposal: The proposed version only
• Original: The original version only

Accept all

Click Accept All to apply all pending proposals at once. If the batch includes space deletions, a confirmation dialog warns that deleting a space removes all its documents permanently.

After acceptance, a confirmation message appears in the chat showing how many changes were accepted and listing the affected documents.

Iterating with AI

When reviewing a proposal from a meeting or outcome, you can chat with AI to refine it before accepting. The detail view opens with a split layout: the proposal on one side, an iteration chat on the other. Describe what you want changed and the AI updates the proposal in place. In meetings, refinements update the same proposal silently, so no new "Capturing..." card flashes back to the top of the meeting.

Each iteration creates a new version. Use the version dropdown to compare versions ("Version 1", "Version 2 (Latest)") and accept the one you prefer.

Some AI actions skip manual review. During meetings, periodic wrap-ups (tagline, summary, screenshot selection) are auto-approved so the event stays up to date in real time. Document creates from background processing may also auto-approve.

Updates, deletes, and space changes always require your approval, regardless of context.

The changes panel on any document shows a chronological history of edits. Each entry shows what changed, who made the change, and when. Click an entry to see details.

Changes are grouped by time and can be expanded to see individual edits. Change types are labeled as Created, Updated, or Deleted with color-coded indicators.

Version history is read-only. It does not currently support reverting to a previous version.

Tana supports models from OpenAI, Anthropic, and Google. Switch models at any time during a conversation. The change takes effect on the next message.

Models are grouped by cost:

• Cheap: Smaller, faster models for simple tasks
• Moderate: Balanced cost and capability
• Expensive: Most capable models for complex reasoning

Some models support reasoning with configurable effort levels (low, medium, high). Some support web search. Both capabilities are indicated in the model selector.

The chat toolbar has two voice modes:

• Dictation: Click the microphone button to speak your message. Words appear in the input as you speak. Supports English, Norwegian, and many other languages.
• Audio recording: Click the voice button to record an audio clip, preview it, and send as an attachment. The audio is automatically transcribed.

A skill is a set of AI instructions packaged for reuse. Instead of typing the same prompt every time you want the AI to do something specific, you write the instructions once as a skill and invoke it whenever you need it. Skills can have parameters for dynamic inputs, making them adaptable to different contexts.

Skills are how you customize what the AI does for your team. A product team might create skills for "Write a PRD from meeting notes" or "Create a GitHub issue from this bug." A consulting team might create "Generate a client summary" or "Draft a follow-up email."

Select "Skill" from the create menu. A skill has four parts:

• Title: The name that appears in search results and toolbar buttons
• Description: When and why to use this skill. Helps the AI decide whether to use it and helps teammates understand its purpose.
• Parameters: Optional inputs the skill needs at invocation time (see below)
• Body: The full instructions, written in a rich text editor with formatting support. This is the prompt the AI follows when the skill is invoked.

Skills live in spaces and are accessible from that space's context.

Skills can be invoked from several places:

• Chat: Type / in the chat input to search and attach a skill. The AI follows the skill's instructions when responding.
• Type toolbar buttons: Skills attached to a type appear as buttons in the toolbar of every document with that type. Click the button to run the skill on that document.
• Meeting outcomes: Choose "Use Skill" from the outcomes panel to run any skill with the meeting transcript and context.
• @mention: Type @ in chat and select a skill to attach it.

Beyond the default Tana assistant, you can create custom agents with their own tools, settings, prompt, and schedules.

Open a custom agent and click Customize next to the title to configure:

• Tools: Pick which tools the agent can use, either by enabling whole bundles or by choosing individual tools
• Settings: Model, temperature, and other generation parameters
• Context fields: What the agent reads on every invocation
• Prompt: The instructions the agent follows, edited inline. An Improve Your Agent helper opens a chat that can read and rewrite the prompt with you.
• Schedules: Pick days and times for the agent to run automatically. Each schedule can carry an optional Additional prompt to vary what the agent does on different runs.

The agent page also shows the agent's recent outputs and chats so you can see what it has been doing.

Built-in agents show a System agent badge instead of Customize. They are managed by Tana and cannot be edited.

The AI can also propose changes to a document's access: restricting it to specific participants with editor or admin roles, opening it back up to everyone in its space or event, and toggling the Guest Link (anyone with the link can view). Access changes go through the same proposal approval flow as the rest of the AI's edits, and the proposal preview shows a clear access callout before you accept.

Every organization picks a default content language in Settings, Organization, Language. The AI writes its replies, generated artifacts, outcomes, summaries, and document titles in that language no matter which language the conversation is happening in. Meeting transcription stays multilingual (auto-detected per speaker) so a mixed-language call still transcribes everyone correctly.

Each organization has monthly and rolling short-window AI usage limits, plus a shared org-wide pool. As a bucket approaches 80 percent, an amber "Approaching usage limit" callout appears at the top of every chat panel and a soft warning lands in Activity. When a bucket is exceeded, the AI returns a red "Usage limit reached" callout instead of a reply. Your own monthly and 5-hour caps are visible as progress bars in Settings, User, Usage.