tutti

--- ## Two Versions of Tutti | Capability | Tutti · Open-source | Tutti · VM | | --------------------------------- | ------------------------------------------------------------------- | ---------------------------------------------------------------------- | | Connect existing agents | ✓ | ✓ | | Use existing subscription credits | ✓ | ✓ | | Built-in Apps | ✓ | ✓ | | Group chat | — | ✓ | | Work with others | — | ✓ | | Work with others' agents | — | ✓ | | Simultaneous editing | — | ✓ | | Agent borrowing | — | ✓ | | Across devices | — | ✓ | | No-deploy sharing | — | ✓ | | Big @ | Across your own agents | Across others' agents | | Download | [Download Tutti · Open-source](https://tutti.sh/en?tc=25q#versions) | [Download Tutti · VM](https://tutti.sh/en?tc=25q#versions) | | Entry | [Straight to GitHub →](https://github.com/tutti-os/tutti) | [Creating a Room takes an invite code →](https://tutti.sh/en/waitlist) | If you like Tutti, give our GitHub repo a star, fork it, open an issue, or send a PR. Anyone interested is welcome to join our [Discord](https://discord.gg/UUemKEWtw6) to share feedback, ask questions, and help shape the future of human-agent collaboration. **Tutti is now open-source.** **Tutti · VM is available during Early Access. Only creating a Room requires an invite code:** **[Explore and download Tutti · VM →](https://tutti.sh/en?tc=25q#versions)** ## Quick Start [Download Tutti · Local for macOS](https://tutti.sh/desktop/download?platform=macos&arch=universal&format=dmg) [Download Tutti · Local for Windows (x64)](https://tutti.sh/desktop/download?platform=windows&arch=x64&format=exe) [Download Tutti · VM for macOS or Windows](https://tutti.sh/en?tc=25q#versions) Creating a Room requires an [invite code](https://tutti.sh/en/waitlist); downloading Tutti · VM or joining someone else's Room does not. See [CONTRIBUTING.md](CONTRIBUTING.md) for the full development guide. ## What is Tutti? Claude Code is powerful. So is Codex. So is Canvas. So is Claude Design. But the moment a real workflow needs dependencies and handoffs, the busiest person in the room ends up being you. Claude finishes an API; now Codex has to build the frontend. So you paste the API docs, bring Codex up to speed on where things stand, and re-explain why you built it that way. And before the frontend even starts, if you want decent UI, you need design and assets too. So you summarize everything again, generate visuals in an image app, then download, upload, and paste them into the next agent and explain the requirements one more time. Agents were supposed to do the work for you. Instead, you've become the messenger between them. ### Tutti provides a real-time shared workspace where context, files, apps, and tasks are all connected Codex can pick up Claude's output without losing any context, as if they shared one brain. On top of that, Tutti has its own app ecosystem: image generation, UI/UX design, docs, and presentations. You can use them, and so can your agents. When Codex calls the prototype-design app, it's as if it gained Claude Design's abilities, and Claude Code can take the result straight into frontend work, with no copy-pasting on your part. **Everything in Tutti is visible and connected. Any output, including what apps generate, can flow between agents and be used directly in the next step.** ## Sounds like you? - You work with several AI agents at once (Codex, Claude Code, Canvas, and so on). - You've copied context between agents more than once, maybe even built your own Markdown-based handoff workflow. - You try to let AI handle everything, but it never quite flows, and every new session starts from scratch. - You've subscribed to a few different AI products, but it never feels worth the money. - For anything complex, your tools sit in separate silos, and the back-and-forth only grows. **Tutti is not a replacement for your coding agent. It's a real-time shared workspace for your agents.**

## Three core features ### 1) A real-time shared workspace Agents no longer just hand off summaries. They share the same real-time workspace, with context, files, running tasks, and apps all connected. Your Codex can see what Claude changed, what is running, and the current state of the project. This unlocks three core capabilities: #### Big @ - In Codex, you can @ past conversations, files, app invocations, and tasks. No repeated copy-pasting or uploading. - You can also @ Claude Code's past conversations, files, app invocations, and tasks from inside Codex, then build on top of them without manually carrying over context. - You can also ask Codex to direct @ Claude Code (app) to do the work.

#### Reference with "+" In the agent chat box, click "+" to reference local files or outputs generated by apps.

#### Task orchestration and multi-project work With shared visibility into each other's work, agents can avoid or resolve conflicts and decide whether work should run in parallel or in sequence. Agents from different providers, such as Claude and Codex, or OpenClaw (DeepSeek) and Hermes, can work together without getting in each other's way. **In Tutti · VM:** - @ works across collaborators too. @ a teammate to pull in any of their agents' conversations, files, or tasks, and click "+" to reference outputs from apps they've used. ### 2) Apps that both you and your agents can use A full workflow is rarely one agent's job. Building a page might start with a prototype, then code, then visuals. Writing an article might need images, layout, and export. Each of these is handled by a different powerful agent or tool, so you end up subscribing to several and switching between them all day: opening, downloading, uploading, screenshotting, pasting. The work itself doesn't get harder; moving information around is what wears you out. Tutti has its own app center, shared in real time across the whole workspace. You can use these apps yourself, or have your agents call them. **For example:** - In Codex, @ a prototype-design app to generate UI drafts, which effectively gives Codex Claude Design-like abilities. The output goes straight into Codex's development work. - Generate visuals yourself in the image app (AI Canvas), then have Claude Code or Codex drop them into the page. - Once you've agreed on an article's structure, @ Codex with a doc app to draft and organize it, then export to HTML. - Need a deck in a few days for an external product intro? @ Claude Code with the AI PPT app to generate one. Want to tweak a few details by hand? AI PPT lets you drag modules around and edit text freely. Every app output stays in the same workspace. When the next step needs it, just reference it with "+" and keep going. Apps also run on your existing agent subscriptions, instead of bundling and reselling capabilities on top of a model. Use official apps, community-built ones, or build your own. ### 3) Less work about work #### Goals to tasks No manual breakdown, no planning every step. Just describe the goal, for example "I want to build a website," and Tutti breaks it into clear sub-tasks. You review them and assign each to the right agent.

#### Control Center No more tab-switching. One view for everything: every agent conversation, every action waiting on your approval, every task that's running. Anything that needs you is surfaced clearly, so you can find it fast and handle it in one click.

#### GUI Interface No command line. Open Tutti and start using agents, apps, tasks, and files right away. Power users skip several steps and cut the friction; builders, designers, and content creators who'd rather not touch a terminal can just get going. ## Reuse your existing subscriptions Connect the agent subscriptions you already have. Every app and agent runs on top of them, at zero extra cost. (Currently Claude Code, Codex, and Hermes; OpenClaw is in development.) ## Tutti vs Tutti · VM | | Tutti (open source) | Tutti · VM (Early Access) | | ----------------- | -------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Best for** | One person, multiple agents | One person, multiple agents
One person, multiple devices
Two or more people, each with their own agents | | **Runs on** | Agents run locally, and the working state stays local | Each agent runs on your own machine inside a managed local VM, while the Room shares working state in the cloud in real time: what's being discussed, what's running, and what's already been created. This enables cross-device, multi-user collaboration among agents, with no context loss, like a shared brain. | | **Shares** | Context, apps, outputs, tasks, and running state across multiple agents | Everything in the local version, plus sharing across multiple people and devices