Connect Provider Accounts

Provider accounts let Apprentice authenticate with the model providers you want to use.

Some providers use OAuth or a local CLI login. Others use API keys. Local runtimes may not need credentials at all.

Open AI Integration

Open Settings > AI Integration.

Use this area to enable providers, configure provider settings, test connectivity, and manage provider accounts.

Account And Credential Types

Apprentice supports several setup styles:

• OAuth or CLI login for providers such as Claude Code, Codex, Gemini CLI, and ChatGPT Subscription.
• API keys for direct API providers such as Anthropic, Gemini, OpenAI, DeepSeek, Mistral, Kimi, GLM, and Qwen.
• Optional or no authentication for local runtimes such as LM Studio, Docker Model Runner, and Ollama.

The exact fields depend on the provider.

API Key Providers

For API providers:

1. Create an API key in the provider's own dashboard.
2. Paste it into the provider settings in Apprentice.
3. Confirm the base URL if you use a custom endpoint.
4. Save and test the provider.
5. Select the provider and model on the agent.

Use separate keys when you want cleaner provider-side usage tracking or revocation.

CLI And Subscription Providers

For CLI or subscription providers:

1. Install or enable the provider's required CLI or login flow.
2. Run the setup command shown by Apprentice.
3. Confirm the provider is detected.
4. Choose a model.
5. Test with a simple agent message.

CLI providers run through the agent runtime path. Do not assume host shell setup alone is enough if the agent runtime still needs authentication state.

Local Runtime Providers

For local runtimes:

• Start the local model service first.
• Confirm the server URL in Apprentice.
• Make sure a model is loaded or available.
• Test the provider before creating a scheduled or integration-triggered agent.

Common local runtimes include LM Studio, Docker Model Runner, and Ollama.

Provider Accounts On Agents

An agent can use a provider default account or be pinned to a specific provider account when the provider supports accounts.

Use account pinning when:

• Different agents should use different billing accounts.
• You want one agent tied to a specific OAuth login.
• You are separating personal, work, or test usage.

Changing an agent's account can retire provider conversation continuity so the next run starts cleanly with the new account context.

Keep Credentials Healthy

If a provider starts failing:

• Re-test it in Settings > AI Integration.
• Check whether the key expired or was revoked.
• Re-run the provider login flow if OAuth expired.
• Confirm the model still exists or is available to your account.
• Check provider rate limits and quota.

Safety Notes

Provider credentials are sensitive. Do not paste keys into agent instructions, chats, memory, or docs.

Use provider dashboards to rotate or revoke keys when needed.

For local runtimes, remember that local inference can still access mounted files only through the agent's configured runtime and permissions.

Next Step

After accounts are connected, set up local runtimes if you want models running on your own machine.