Collaborators
Invite teammates to chat with the agent on the same project.
Collaborators are people who edit your game with you. They can chat with the agent, drag entities in Edit mode, tune Inspector fields, and publish — same access as the owner, minus member management.
When to use collaborators
- You're working with a co-creator — design + code, two devs, etc.
- You're teaching someone — pair-edit a game together.
- You're handing off work — let another person continue the project while you step away.
If the other person should only play (not edit), invite them as a Player instead. See Invite players.
Who pays for a turn
Agent turns in a shared project are billed to one account per editing session: the account that opened the project's session, which is whichever of you got there first after it had been idle. Every turn run in that session is charged to that account, whoever typed the prompt.
In practice that usually means the owner, because the owner usually opens the project first — but not always. If it matters to you, open the project yourself before your collaborator does, and check Settings → Credit Usage, which shows every deduction with the game it belonged to.
There are no per-collaborator quotas today. The model is trust plus visibility: anyone with editor access can spend, and every turn is on the ledger.
Inviting a collaborator
Same UI as players, different role:
Open the Members tool — the ≡ menu in the editor, or the tab strip's +.
In the invite form, type a username or email.
Pick Collaborator in the role dropdown.
Click Send, confirm. The invitee gets a notification (signed-in accounts) or branded HTML email.
What the collaborator sees
After accepting:
- The project appears in their Projects page with a Collaborator badge.
- They can open it like any of their own projects and chat with the agent.
- They share the same workspace — the same files, the same chat history, the same edit-mode state.
You can both be in at once
Multiple collaborators can have the project open at the same time, live. There is no lock and no one gets bumped — opening the editor while a teammate is in it is fine.
What's shared, as it happens:
- The chat. Their prompt appears in your chat the moment the agent starts on it; so do the agent's replies, its activity, and the new preview build.
- Who's here. An avatar stack in the header shows who else has the project open right now.
- The queue. See below.
Your own second tab still displaces your first — that part is intentional, so one person's duplicate tabs can't fight each other.
One turn at a time, and a shared queue
The agent runs one turn per project. Anything sent while it's busy is queued on the server, not rejected and not held in your browser — so two people typing at once cannot collide or silently lose a message.
The queue is visible to everyone in the project, in order, with who sent each item, and any of you can drop an entry that's no longer wanted.
There is also a Stop button while the agent is working. It interrupts the current turn cleanly rather than killing it — useful when someone realises mid-turn that they asked for the wrong thing.
Coordinate anyway. Two people steering the same agent in opposite directions is a people problem the queue cannot solve.
Removing a collaborator
In Members → Collaborators, hover and click the X. Confirm.
Removed collaborators:
- Lose editor access immediately.
- Keep their umicat account and any other projects they own or collaborate on.
- Their previous chat / edits remain in your project (the work is yours; the access is theirs).
Leaving a project you collaborate on
If you collaborate on someone else's project and want out, open the project → Members → Leave. A confirm modal appears. Leaving revokes your access; the project is unaffected.
The owner cannot accidentally remove themselves — leaving is for non- owners only.
Pending invitations
Like player invites, pending collaborator invites show in Members with revoke buttons. They expire after ~7 days.
What collaborators can't do
| Action | Permitted? |
|---|---|
| Chat with the agent | ✓ |
| Edit in the canvas | ✓ |
| Drag-to-place from Assets | ✓ |
| Use the Inspector | ✓ |
| Publish | ✓ |
| Manage members | ✗ (owner only) |
| Delete the project | ✗ (owner only) |
| Change project metadata (title, description, tags, orientation) | ✗ (owner only) |
| Unpublish the game | ✗ (owner only) |
| Make the game forkable | ✗ (owner only) |
| Transfer ownership | ✗ (owner only — and not yet supported anyway) |
Use case examples
Two-person team
You're the designer, your friend is the gameplay-tuning person. They
collaborate on the project; you both chat with the agent over time;
you sync on Discord about who's doing what.Pair learning
You're showing a friend how to use Umicat. Add them as Collaborator.
You both have the editor open at the same time, watching the same preview
rebuild — they type a prompt, you both see the result.Async handoff
You start a project, get tired, hand it off to a collaborator who
continues for a week, then hands back. The chat log shows what
happened while you were away.