The new Tana opens to the community

The new Tana is live for community members. Not a public launch. A community launch. The product is still moving fast, changing daily, and not yet where we want it to be. But you can feel the magic now, and we wanted you to be the first to try it.

We've always built Tana with the community. When we started, there were four people in the community and we met with them every other week. That hasn't changed. This is a feedback round: use it for a few days, do your meetings there, build your first agent, and tell us what works and what doesn't.

We've deliberately launched early so we can start exploring and refining the product. That means you'll encounter rough edges. We know. Some parts of the product haven't received focused attention yet, and others are actively being rebuilt. Bugs and feedback on the current state are welcome and expected. The T2 channel in Slack is the place for it.

For users coming from Tana Outliner

Three key differences:

Block editor, not outliner. A modern block-based editor with headings, tables, embeds, and rich content, better for long-form writing.
AI configures things for you. Instead of manually setting up supertags, fields, and views, describe your workflow and the AI builds the structure.
Types replace Supertags. The concept is preserved. Supertags are now called Types, and Fields are called Attributes. The mental model carries over, now with AI doing the heavy lifting.

Meetings as work sessions

Meetings are the most information-dense place in a company. That's where decisions are made, where context is shared, and where the most important discussions happen. The new Tana treats meetings as a working surface, not just something to record and summarise.

The meeting experience has been a major focus of the Tana 2 team. The tile system, the navigation, the overall experience of being in a meeting have been where much of the effort has gone. It's the place where you get the most distinctive sense of Tana 2's new identity and what the product is becoming.

In practice, this means the AI is present during the conversation. It listens, captures, and acts. You can ask Tana to find a document and pin it to the meeting, surface your open tasks related to a client, or pull up notes from the last time you met with someone. All by voice, mid-conversation. No searching, no tab-switching.

The live meeting digest

The live meeting digest has a status line from the first second of a call: "Waiting for conversation…" before audio, "Following Along" once words arrive, and "Thinking…" while extracting a new topic section. Status is broadcast to every participant. Topic sections start appearing after roughly thirty seconds of transcript. Every call ends with a structured summary (Key Takeaways, Topics Discussed, and Next Steps) written and pinned automatically.

A few things you can do inside a meeting:

Voice-driven context. Ask Tana to find and pin documents, tasks, or notes to the current session. It searches your workspace and brings what's relevant in, so the information is there when you need it.
Storyboards and capture. The AI watches what's being said and, optionally, what's on screen. You can ask it to create a storyboard from an anecdote, capture a quote for later use, or generate a summary of a specific segment of the conversation. You decide what gets pulled out and how.
Screen sharing as context. When you share your screen, Tana sees it. The AI can reason about what's shown alongside what's being discussed. Useful for handing off issues to a coding agent or capturing annotated screenshots with relevant context attached.
Canvas. A fully collaborative infinite canvas, available inside any meeting. Draw, write, map out strategy together in real time. The AI has access to it too, and can base its outputs on what's been drawn or written.
Solo sessions. These don't have to be group meetings. A solo session with Tana is just as powerful: talk through a slide deck before a board meeting, wake up on Monday and ask what happened last week, or reflect on your day with a personal agent. Olav uses a daily reflection agent that reviews what he planned, what he did, and where things stand. The first time feels like magic. The third time, it's just how you start your day.

Voice agent and "Hey Tana"

Tana has a voice agent activated by saying "Hey Tana" in any meeting, with wake words in nine languages. No new interface to learn, just talk. The voice agent control lives in the meeting toolbar action pill. When two participants say "Hey Tana" simultaneously, whoever starts first keeps the agent; the other must explicitly unmute to take over. The voice agent tile uses a shader-based animation that reacts to speech with motion across three brand colors.

Guest experience

Guests who open a meeting link see the full meeting experience (live digest, pinned docs, and approved outcomes) right from the link. Before the host starts, guests wait in a lobby and join automatically when a Tana user begins the call. Cross-org guests see a "Join meeting as guest?" prompt instead of a dead end.

Capturing external meetings

Tana captures meetings from Google Meet, Teams, and Zoom. From the Today page, calendar event cards show a "Capture Meet", "Capture Teams", or "Capture Zoom" button when Tana detects a matching meeting on the user's computer. Captured external meetings auto-generate a tagline, summary, and outcome proposals when the recording ends, with no manual step needed. A live screen-share tile updates in real time during capture. Transcript, screen shares, and action items all flow into the Tana workspace automatically.

Hand off to coding tools

Context from a meeting or chat can be sent directly to Claude Code, Cursor, Codex, Lovable, or v0. Discuss a feature in the meeting; the work starts moving in your engineering tool of choice.

Shareable artifacts

Storyboards, customer journeys, and slides can be shared via public link with all images and layout intact. These are fully rendered artifacts produced by the AI from meeting content, not ephemeral chat messages, but lasting knowledge artifacts you can share and act on.

Default content language

Set it once, and the AI writes in your language regardless of what's spoken on the call. Non-English organizations get AI output in the configured content language across all surfaces: live meeting digest headings, post-meeting tagline, summary, Key Takeaways, Topics, Next Steps, screenshot alt-text, annotation labels, and auto-titles on uploaded files. Meeting transcription itself stays multilingual.

AI-managed workflows and types

Ask the AI to add a workflow to a type, rename or reorder states, add new states like "Blocked", or remove a workflow entirely. Each change goes through the standard proposal-approval flow. Renames preserve a state's identity so existing items keep their progress. When the AI proposes creating a typed item in one space but the type is scoped to a different space, the proposal shows a yellow Location callout and approval places the item in the type's home space. Post-meeting outcomes route to the right type: a user observation can be classified as Feedback, Friction, or Pain Point instead of defaulting to Bug.

Over the weeks and months ahead, more effort will expand to other parts of the Tana 2 feature set. Tasks, people, hierarchy and overview, navigation, and more will start to become more aligned with the vision that meetings have established. The meeting experience is the leading edge, and the rest of the product will follow.

Building agents

Agents in the new Tana let you automate repeatable work by describing what you need in natural language. They can use skills and tools, run on a schedule, and operate across your meetings, documents, and workspace.

In May's Tana Current, Brage gave a demo of his meeting prep agent that runs every morning at 8 a.m. It goes through his calendar, checks who he's meeting with, reviews his past meeting history with each person, pulls their HubSpot record if they're a client, and generates a prep document with talking points, key context, and relevant pinned documents, all attached to the meeting before it starts.

Creating an agent like this takes minutes. You describe what you need to the AI in the agent creator, review the proposed workflow, and set it to run on a schedule. Adjusting how it works is the same process: talk to the AI again and tell it what to change.

If you have repeatable tasks, whether that's prepping for meetings, updating project tracks after a call, or running a weekly review, you can build an agent to handle it. You can create them in the general AI chat or through a dedicated agent creator, and they can be triggered manually or scheduled to run automatically.

The team are actively working on the final stages of our first MCP connection tool that will allow you to bring in other tools and connect to them during your meetings. This is going to be a dramatic shift in the potential and use of T2.

The collaborative knowledge graph

Apart from a new meeting experience, what makes the new Tana different from a meeting recorder is the knowledge graph, or context graph, as some are calling it.

Every meeting, every document, every captured item feeds into a shared graph that the AI can traverse. This is what allows it to answer questions like "what are we working on today?" or "what did we discuss with this client last time?" with real answers drawn from your actual history.

The key innovation is branching: like git for knowledge workers. When the AI proposes changes based on a meeting (updating a project status, filing a new issue, revising a strategy doc), those proposals go through review before they enter the graph. You see what the AI wants to commit, and you approve, edit, or reject it.

You may not feel the importance of this with a handful of documents. But at scale, with a million docs, it becomes essential that the graph contains things you actually need and that the AI isn't traversing noise. The review step keeps the graph clean. Directing the AI to specific spaces and points in the graph also keeps token costs manageable.

A new identity

Tana Outliner and the new Tana are separate products that are both important to the team. Each has its own team, its own roadmap, and its own future. We wanted to keep the Tana name for the new product largely because it represents our team and everything we are. But we recognise that sharing a name introduces overlap with the existing Outliner.

The way the team planned to address this was by creating a distinctive visual identity for the new Tana, while keeping the original Tana branding with the Outliner. Tana Outliner retains the existing logo and identity it's always had. The new product gets something new.

Mathias Hovet, a Norwegian designer, was brought in to lead the new identity. He got an introduction to the product, tried it, and landed on the metaphor immediately: a prism. Everything you need is already in your conversations. The decisions, the context, the next steps. The new Tana pulls them out, separating the clear light into colors. That's the role it plays, and the prism captures it.

The new logo, colour palette, and visual language will carry across the app, marketing, and content. When you see the prism, it's the new Tana. When you see the original logo, it's the Outliner. Two products, two identities, clearly separated: in the app, in content, on YouTube. You'll start to see this reflected across our channels in the coming weeks.

The brand story: ending meeting debt

The core brand narrative centers on "meeting debt", the gap between the meeting and the work. Work went remote, then hybrid, then async. Organizations added Zoom, Slack, Notion, Linear, dozens of tools to stay connected. But the fundamental problem got worse: the meeting and the work became two separate things. Teams meet to decide, then scramble to document the decision. They meet to plan, then rush to write the plan. The meeting ends, the work begins, except there's no time left to do it. Decisions get re-litigated, action items forgotten, context lost across tools and time zones.

The tagline: Tana is the prism for your work. It joins meetings the way teams already work, through video calls and natural conversation, with no new interface to learn. While you talk, AI listens, understands, and acts. What comes out isn't added complexity. It's the same conversation, separated into its true colors: the insight buried in the discussion, the action nobody wrote down, the decision that would have saved three follow-up meetings. Tana reveals what was always there, separated into clarity.

The vision statement: we're entering an era where talking about the work equals doing it. Your team discusses the feature, the PRD writes itself. The call ends, the follow-up is drafted, the CRM updated, the tasks filed. Intelligence refracts from conversation to finished work without the scramble in between.

Domain and product separation

On March 16, 2026, the team decided to claim tana.inc for the meetings product and move the outliner to outliner.tana.inc. The meetings product owns the original Tana brand and app icon. The outliner gets a white icon with "Outliner" branding. This was timed for GTC week in San Francisco, so visitors to tana.inc see the new meetings-focused messaging. Two products, two identities, one company.

What's coming next

The product is still early. The editor needs more work. Task management is being rethought. The external MCP server is ready but being optimized so users don't burn through credits in ten minutes. When it ships, it will bridge the two products. You'll be able to connect your own Claude agent or any external tool.

Updates are shipping daily. The pace from here only accelerates.

Tana Outliner

The Outliner team has its own roadmap and its own momentum. Q2 remains focused on improving the everyday experience: core reliability, paper cuts, and long-requested features.

Recently shipped

Set fields by dragging in grouped outline view
Backup audio recording for the meeting notetaker. Retranscribe from audio files or in different languages after network drops
Copy node ID command for MCP workflows
Formatting preserved when pasting code
Improved autocomplete for fields with thousands of options
Several MCP fixes, shipped continuously

New documentation site

A major overhaul of the docs and help center, led by Theo. The new site is built closer to the codebase, so documentation updates more automatically when features ship. It includes better search, feedback buttons, and a related content section. The site is also built for agents as well as humans: asking Claude or the built-in Tana helper about Tana should now return better, more accurate results.

Semantic field breadcrumbs

A big one for power users. The semantic function lets you map relationships between nodes, for example: Oslo is a city in Norway, which is a country in Europe. The first half of this feature (the underlying function) shipped earlier. The second half, the visualization, is now landing. Navigate to any node and see breadcrumbs showing its place in the semantic hierarchy. Useful for topic maps, feature hierarchies, geographic relationships, and more. Set it up in the field config under the advanced section.

Mermaid diagrams from AI chat

You can now ask for mermaid diagram output from the AI chat and get rendered diagrams directly in Tana. This has been in internal testing and is shipping to the community in the next few days. Feedback welcome.

Coming next

Collapsible fields. Much requested, tricky to get right, but close
Continued grouping improvements
Startup speed improvements for the desktop app
MCP improvements, a long list of smaller fixes
Quality of life improvements from community feedback
Option to opt out of in-app popup guides
Ability to hide empty related content

Hiring

The Outliner team is adding a senior software engineer, ideally based in Sydney, Oslo, or Trondheim, for proximity to existing team hubs. If you know someone or are interested, the role will be posted soon.

Community events

Mark's MCP Workshop continues every two weeks. The focus is on thinking in terms of processes and objects, which lets you build workflows with the MCP that can route things and give you real power. It's not just the techniques of Tana in these workshops, it's how to build systems and how to do them well. Always recorded and shared in the community.

MCP Help Sessions run every second week: unscripted, hands-on help with setup, troubleshooting, and workflow design. Two sessions are scheduled for next week. Check Luma for details. The MCP setup is straightforward once you've done it once. If you haven't, these sessions are the fastest way in.

What's ahead

The new Tana is in the community's hands. It's early, it's changing fast, and it's not finished. But the direction is clear, and we're building it together.

The new Tana is live for community members. Use it, break it, tell us what works
Agents, voice commands, and the knowledge graph are the core of the new product
The external MCP is soon ready and will bridge the two products when it ships
Tana Outliner has a dedicated team shipping improvements every week: semantic breadcrumbs, mermaid diagrams, a new docs site, and more on the way
We're hiring on both products

Keep sharing feedback and ideas in the community, it's where everything starts.

Watch the live recording on Youtube

Next Tana Current: to be announced in Slack

Tana Current May 2026