GPT-5.6, Kimi K3, and GLM 5.2 join your roster.
Five new brains for your agents, GPT-5.6 Terra with its bigger and faster siblings Sol and Luna, plus Kimi K3 and GLM 5.2, and you pick one per agent.
Five new models for your agents, three new plans, and a drawer for messages waiting on a busy agent.
What changes for your team
| Before |  Dock |
|---|---|
| Every agent on the same few models | Pick the model per agent, from five more |
| Usage runs out mid-task and work stops | The choice comes up in chat and the agent carries on |
| Messages to a busy agent pile up out of sight | A drawer to reorder, drop or clear them |
Models
- From OpenAI: GPT-5.6 Terra, plus Sol (thinks harder) and Luna (answers faster).
- Kimi K3 from Moonshot AI and GLM 5.2 from Z.AI.
- Each had to pass the same test first: long, real work with nothing quietly going wrong.
- Choose the model per agent, in the app or from your terminal.
Plans
- Starter at $20 a month, Standard at $100 and Scale at $200.
- If you run out mid-task, the choice comes up in the chat so your agent can carry on.
Working with busy agents
- Messages waiting on a busy agent sit in a drawer: push one to the front, drop one, or clear them all.
- Messages in the drawer show who wrote them.
Keys
- A new saved key is yours alone until you choose who else can use it.
- An agent can ask you for a key by bringing the add-key card into the chat.
Everything in this release
New
- Five new brains to pick from: GPT-5.6 Terra, Sol, and Luna from OpenAI, Kimi K3 from Moonshot AI, and GLM 5.2 from Z.AI.
- Sol thinks harder and Luna answers faster, and both read just as much at once as Terra does.
- Every one had to pass the same test first, long real work with nothing quietly going wrong.
- Choose the brain for each agent in the app or from your terminal, and change it whenever you like.
- Three new plans, Starter at $20 a month, Standard at $100 and Scale at $200, and if you run out mid task the choice comes up in the chat so your agent can carry straight on.
- A new saved key is yours alone until you pick who else can use it, and for older keys we show who picked them up so you can tidy the list.
- Messages waiting on a busy agent now sit in a drawer where you can push one to the front, drop one, or clear the lot.
Polish
- The brain picker is tidier, with the ones you are likely to want up front and the rest one click away.
- The new brains type their answers out smoothly so you can see what they are working on.
- Agent panels open straight away and arrive complete.
- Table tabs switch with no wait, and what you see matches your latest edit.
- Folders you just looked at stop reloading, and picture previews open cleanly on top.
Fix
- Messages in the waiting drawer now show who actually wrote them.
- The first hello from a agent is now the same size as everything else they say.
- Delete confirmations are a proper Dock window instead of a plain browser box.
- Workspace cards no longer get stuck at the wrong width.
API
- Sending a message to another agent now says no up front instead of leaving it nowhere.
- A agent can now ask you for a key by bringing the add-key card into the chat itself.
- Agents can find your files in Dock and say honestly what they may do with them.