Connections (My Settings → Connections) lets you attach model endpoints for your own use — your local model service or an external provider you are allowed to use. Models from your connections then appear in your model selector. Workspace-wide connections managed by administrators are covered separately in Connections (Admin).
Add a connection
Open the Connections section
My Settings → Connections → Add Models.
Choose the endpoint type
Options typically include:
- Ollama / Local Ollama — a model service on your machine or network.
- OpenAI Official — the vendor's hosted API (subject to your organisation's policy).
- Custom / Custom Endpoint — any OpenAI-compatible server.
Provide details
Enter the connection URL (required), any API key the endpoint needs, and additional configuration where offered (the configuration must be valid).
Test then save
Use the test action — a successful test confirms the endpoint is reachable and credentialed before you save.
Manage connections
The list shows each configured endpoint; select one to edit or remove it. If nothing is configured the page notes "No connections configured".
Rules and etiquette
- Personal scope — your connections affect only your account.
- Policy first — using external endpoints sends prompts to that service. Follow your organisation's rules; administrators can disable external connections entirely. See Connections (Admin).
- Protect keys — never paste provider keys into chats or documents; store them only in this settings page (or your password vault).
- Costs — external usage is billed by your provider unless the administrator negotiated otherwise.
Troubleshooting
| Message | Meaning / fix |
|---|---|
| "Connection URL is required" | Enter the endpoint address. |
| "Config must be valid JSON" | Fix the configuration syntax you pasted. |
| Test fails | Wrong URL/key, network unreachable, or the service isn't running. |