Models
Connect your ChatGPT plan or an API key, choose which AI model each agent thinks with, and tune how hard it reasons.
Overview
Every agent runs on an AI model from an AI provider (ChatGPT, OpenAI, Anthropic, Google, …) you connect. Opereo hosts the agent; the thinking happens on your own plan or key, so there are no per-message fees from Opereo.
What you can do
Sign in once with your ChatGPT plan or paste a provider key, then switch any agent between the models that provider offers — heavier for hard work, lighter for quick replies.
An agent can't answer until at least one provider is connected. Connect a model is the first item on the setup checklist for that reason.
Open the models settings
Open the agent's settings (the gear on its card, or Employee settings in the chat header) and choose models. You'll see:
- A Run on your ChatGPT plan banner if ChatGPT isn't connected yet.
- The Active model — the one this agent is using now.
- The Models list — everything you can switch to, from every connected provider. Search it, filter by provider, or sort.
- Providers (top right) — every provider Opereo can connect, with a dot showing whether it's connected.
- The Model behaviour gear — Reasoning effort.
Connect your ChatGPT plan
If you already pay for ChatGPT, this is the quickest route and it covers every agent on the workspace.
Click Connect ChatGPT on the banner (or pick ChatGPT from Providers).
Click Sign in. A ChatGPT window opens; sign in and approve.
If the window shows a code, paste it into Opereo where it says Paste the code from that window and click Connect. Otherwise come back and click I've signed in.
Sign in once and every agent on this workspace can run on ChatGPT. Their models appear in the Models list.
Connect a provider with an API key
Use this for OpenAI, Anthropic, Google, and any other provider that hands out API keys.
Click Providers and pick a provider marked API key.
Paste the key from that provider's dashboard where it says Paste your … key, and click Connect.
The provider's models join the Models list for this agent.
Keys are stored on the agent's private workspace and never shown again in full. To remove a provider, open it from Providers and click Disconnect.
Some providers are marked OAuth (you sign in, like ChatGPT). Connect ChatGPT or an API-key provider before sending your first message. OpenCode Free models are unavailable; paid OpenCode Zen models remain available with a connected key.
Switch models
In the Models list, click a model and then Use this model. The change applies to this agent only, from its next message.
You can also switch from the chat: click the model pill next to the composer and pick a different model. Same list, same effect.
Which model should I use?
| Task | Pick |
|---|---|
| Quick replies, summaries, formatting, routine questions | A lighter, faster model |
| Research, planning, multi-step tasks, anything with numbers | A heavier model |
| Long documents | A model with a large context, on higher reasoning effort |
Two agents in the same department can run on different models. Compare by asking both the same question.
Reasoning effort
Click the Model behaviour gear and choose Reasoning effort: how hard the model thinks before it answers. Options run from minimal to max. More effort is slower but more thorough. Leave it in the middle for everyday work and raise it for tasks that deserve a second look.