--- title: Providers, Models & Agents description: Connect AI providers, pick models, and set up agents. --- # Providers, Models & Agents Before OpenChamber can do anything, it needs at least one AI provider connected. This page covers connecting a provider, choosing a model, and tailoring agents. ## Connect a provider 1. Open **Settings → Providers**. 2. Open the **Add provider** menu and pick a provider that isn't connected yet, or choose **Other / Custom** for an OpenAI-compatible endpoint. 3. Sign in one of two ways, depending on the provider: - **API key** — paste your key and save. - **Sign-in (device flow)** — OpenChamber shows a link and a short code. Open the link, enter the code, and approve. OpenChamber finishes connecting on its own. ### Custom / Other providers For gateways, campus LLMs, Ollama, LiteLLM, and similar OpenAI-compatible APIs: 1. Choose **Other / Custom** in the provider list. 2. Enter a provider ID, display name, base URL (`http://` or `https://`), API key (or `{env:VAR_NAME}`), and at least one model id/name. 3. Optionally add request headers. 4. Save — OpenChamber writes the provider block to OpenCode config and stores the key in OpenCode auth (literal keys) or records an `{env:VAR}` reference. 5. To change an existing custom provider, open it and choose **Edit**. Custom providers require an API key or `{env:VAR_NAME}` before they show as fully connected. Without credentials, models may appear but chat calls will fail. When a provider shows as connected, its models become available in chat. To disconnect, open the provider and choose to remove its sign-in. ## Pick a model You pick the model where you're working: - in chat, use the model picker in the message bar to set the provider and model for that session - per agent, set a default model (below) ## Set up agents An agent is a named setup — a model, a personality, and what it's allowed to do. 1. Open **Settings → Agents**. 2. Pick an agent, or create a new one. 3. Edit any of: - **description** — what the agent is for - **model** — its default model - **temperature** — how creative its answers are - **prompt** — standing instructions it always follows - **tool rules** — which tools it may use ## Where your sign-ins live Provider sign-ins are stored by OpenCode, not OpenChamber, so they're shared with the OpenCode CLI. If you set the same provider in more than one place, the most specific setting wins: a per-project setting overrides your personal one. ## Related - [Integrations](/integrations/) — use a Claude, Command Code, or Cursor subscription as a provider - [MCP Servers](/mcp/) — add extra tools for agents - [Usage & Quotas](/usage/) — track how much you've used