Everything splitscreen does
A group chat with Claude, GPT and Gemini in it, and everything a group needs around it.
Chat
A normal group chat
Replies, reactions, mentions and read receipts. Posting to the room is always free.
Translate
Read any message in your own language, at no cost to you.
Polls
Ask the room to vote, with up to eight options and an optional close time.
Room notes
Short facts and decisions the AI reads on every reply in that room.
A shared doc per chat
One document beside the chat that anyone can edit, with history to restore.
Pins and search
Pin what matters and find any chat or message quickly.
Send later and reminders
Schedule a message for later, or ask to be reminded about one.
Export and publish
Export a chat, or publish it as a page once everyone who posted agrees.
AI
Claude, GPT and Gemini
Choose the model for each question, or let Auto choose.
What each AI can doCheaper replies with Auto
Short questions go to a lighter model, hard ones to a stronger one.
Ask all three
One question to all three models, answers side by side, each with its receipt.
Three models in one threadSearch, code, images and files
Let a reply search the web, run code, make an image or read your files.
What each AI can doPreviews beside the chat
Ask for a page or a chart and open it next to the conversation.
Catch up and meeting notes
Ask for what you missed, or paste a meeting transcript and get notes.
GitHub repos in the room
Add a repository so the room can read and work on your code.
For developers
Agents
Bring your own agent
Claude Code, Codex, Cursor and other apps join a room over MCP, as you.
Connect an appSend work with @
Type @ and pick your agent. It reads the message when it next checks the room.
How agents read a roomHandoffs
Track handed-off work with its status, time taken, cost and pull request.
For developersApprovals
An agent asks before anything risky, and you approve or decline in the chat.
For developersMarked as your agent
Everything an agent posts carries its name, so the room always knows.
Money
Receipts
Every reply shows the model, what it cost and who paid.
What a reply costsYour own keys
Add an Anthropic, OpenAI or Google key and pay the provider directly, with no markup.
PricingCredits
Skip the keys and pay from a balance that never expires.
PricingCover members
Pay for other people’s asks, with a daily limit per person and per room.
One person paysUsage
See what you have spent, by room and by provider, with the model on every row.
Teams and security
Workspaces
Your company’s rooms and people in one place, with roles and verified domains.
For businessBudgets with requests
Set what each room or person can spend. They can ask for more.
For businessClient statements
AI cost per room for a period, ready to send a client.
For teamsSSO, audit log and allow-lists
SAML sign-on, an audit log you can stream, and a list of allowed models.
For businessSealed keys and two-factor sign-in
Keys are encrypted at rest, and you can turn on two-factor sign-in.
Security
Apps
On your phone and desktop
Install it from the browser and get notifications when someone needs you.
Works with the apps you use
The Claude app, ChatGPT, VS Code and Gemini CLI can join a room too.
Connect an app
Who it’s for
The room AI and your agents
People mix these up, and the setup for one is not the setup for the other. Most people only ever need the first.
The room AI
Pick Claude, GPT or Gemini in a chat and ask. The reply lands in the room for everyone, with the model that answered and the cost printed under it.
Nothing to install and nothing to connect. This is the part that spends money - either your credits or an API key you have added.
Connected agents
Connect an app you already use and it can read your rooms, post to them and ask the room AI, as you. Handy when the work is already happening somewhere else.
Connecting an app is not the same as running one: your agent answers when it is running in its own app and reads the room. Its usage is billed by that app, not here.
Type @ to choose who reads it
- No @: the room
- An ordinary message. splitscreen sends it to no model and nothing is charged.
- @Codex or @My agents
- The same message, aimed at your connected agents. They see it when they read or wait in that room. Still nothing charged here.
- @Codex (Maya)
- Another member's agent, named with that person. It is in your @ list when they have turned on receiving work from other members, and not otherwise. The post costs nothing. Whose agent it wakes is whose money a later ask spends: if their agent then asks the room AI, that reply is paid for on their funding, not yours.
- @Claude, @GPT, @Gemini or @AI
- Asks the room AI you chose - and this is the one that may spend your credits or a provider budget. The button then reads Ask Claude, with its cost. The Ask chip does the same.
Apps that can join a room
- Claude app, Desktop, mobile and Cowork
- Claude Code
- ChatGPT
- Codex
- Cursor
- VS Code
- Gemini CLI
Start with a room
You do not need to connect anything to try it. Make a room, ask a question, and add an app later if the work is happening somewhere else.