# Live rooms

> How a page updates on every open screen, how many rooms run at once, and what a visitor sees when none is free.

Kind: concept · ~3 min · Verified v0.237.0 · Next: [Secrets an agent uses but never sees](https://pr-2197-stg.superartifacts.app/docs/concepts/secrets.md)

A **live room** sends data changes to a page's open screens without a reload. One artifact has at most one data room, with up to 500 open connections on every plan.

The room carries updates, not writes. Valid answers save through the data API even when a live connection is refused. The page must handle that refusal and read updates another way while waiting.

## How many rooms at once

A workspace's concurrent room limits are Free 1, Solo 3, Team 10 and Business 50. [Limits](https://pr-2197-stg.superartifacts.app/docs/reference/limits.md) gives the current plan table.

- The first connection takes a slot, not publication.
- The last departure starts a 20-second grace period reserved for the same artifact. A silent network loss takes about 105 seconds to detect before that grace.
- Connections renew their data token after 15 minutes. Reuse a still-valid token when retrying capacity; minting one on every retry can hit the token limit.
- Turning live off releases the slot immediately and disconnects its screens.

## When every room is taken

The served header can show **Live paused · input is saved**. The busy response tells the page to retry after 20 seconds; platform live templates handle that retry. Custom pages must implement it too.

**Which artifacts hold a live room.** Your agent lists held rooms and changes nothing until you choose one.

```prompt
Which of my Super Artifacts are holding a live room right now? Call super_whoami and read liveRooms: tell me how many rooms my workspace may hold at once and, for each one held, the artifact and since when. If I want another artifact live and every room is taken, suggest which to turn off with super_artifact_set live false, and change nothing until I say which; afterwards tell me how many open pages it disconnected.
```

Close the other artifact's tabs or ask the agent to turn its live connections off. An open tab can keep the slot.

## Who hears what

Each declared **stream** has a read audience:

| `read` | Visitor socket | Owner-scope socket |
| --- | --- | --- |
| `own` | Its own rows | Its own rows |
| `contributors` | Others' rows after contributing | The same |
| `public` | Readable rows | The same |
| `admin` | Nothing | Readable rows |

Other people's rows remain masked even on an owner-scope socket. A stream declared as an aggregate or count carries no rows. Named signed-in readers of a private artifact can receive another named writer's display name as author.

## The host and the phase

An Owner or Editor can act as host. A **phase**, such as `question` or `results`, is saved before it is announced. Set it through the host connection or `super_artifact_set`. `writeAt` accepts answers only during a phase; `revealAt` opens reads from its phase onward.

## Document editing is separate

Co-editing document source uses a different room and capacity pool: three open document rooms on Free, 20 on paid plans. It does not use the response-room limits above.

## Check it worked

Submit from one browser and watch another. Test capacity refusal as well as a successful connection; writes should still report their actual outcome.

## Next

[Shared pages](https://pr-2197-stg.superartifacts.app/docs/shared-pages.md), [agent setup](https://pr-2197-stg.superartifacts.app/docs/agent-setup.md) and [close codes](https://pr-2197-stg.superartifacts.app/docs/reference/errors.md).
