Configuration
Change an agent's model, provider, memory, and web search from the console Settings tab, and connect its channels from the agent's own dashboard.
An agent is configured in two places after it's deployed: the console Settings tab controls its model, provider, memory, and web search; the agent's own dashboard controls its channels.
Both are stored on the agent itself — set once, and the running agent keeps using them for every conversation, instead of you resending the configuration each time the way you would in the Inference Playground.
In the console: the Settings tab
Open an agent and go to its Settings tab to change how it thinks — its provider, model, memory, and web search:
| Parameter | Type | Description |
|---|---|---|
Agent name | text field | The display name for the agent. |
Provider | selector | The model provider backing the agent — Claude, OpenAI, or Gemini. |
Model | text field | A specific model id; leave blank for the provider default. |
New API key | password field | Rotate the provider key — leave blank to keep the existing one. |
Memory | toggle | Persistent memory across conversations, so the agent recalls prior context. |
Web search | toggle | Allow the agent to use web search for up-to-date answers (requires a plan that includes it). |
Click Save changes to store them on the agent.
In the agent's dashboard: channels
Channels are not in the console — they live in the agent's own dashboard, which you log into with the admin username and password you set when you deployed it. Connect a channel by pasting in its credential (a bot token or API key); how many you can run at once is capped by your plan.
- 1Open the agent's dashboardSign in to the agent's own dashboard using the admin username and password from deploy, and open Channels.
- 2Add the channelAdd the channel (e.g. Telegram or, on OpenClaw, Slack) and paste in its bot token or credential.
- 3Restart the agentBack in the console, click Restart so the running agent connects to the newly added channel — a saved channel does not take effect until the agent restarts.
Switching model or provider
Change the Provider and Model dropdowns to switch what the agent uses to generate responses — useful for moving a support agent from a faster, cheaper model to a stronger one as needed.
Memory and web search
Memory and web search are simple toggles. Turning memory off makes the agent stateless between messages; turning it on lets it draw on earlier conversation history. Web search only appears usable on plans that include it.