Choose your providers

Keep multiple AI providers connected at once: Apple Intelligence, Ollama Local, Gemini, OpenAI API, ChatGPT / Codex, Claude, OpenRouter, Perplexity, DeepSeek, Groq, MiniMax, Claude Code CLI, an OpenAI-compatible API, Ollama Cloud or Temporary AI. Availability depends on your Mac, account and connection. Each action uses the provider you choose; ActionClip does not silently switch providers.

Connect in one place

  1. Open ActionClip → Settings → AI → Configure….
  2. Select a provider in Configure AI providers. For an API provider, enter your own key and choose Connect. For ChatGPT / Codex, use Connect ChatGPT and complete the OpenAI account flow.
  3. Connect other providers without disconnecting the first. A checkmark identifies configured providers.
  4. Choose Done to return to Settings.

A connected API provider shows a masked key ending in its last four characters. Choose Disconnect to remove its saved connection and return to the empty connection form. To replace a key, disconnect and connect with the new key.

Set the default provider and model

In Settings, choose Default AI provider from your connected providers, then Default AI model. The configuration modal also has a Provider default model for each connected provider. Apple Intelligence and Temporary AI choose their models automatically.

Actions using Settings default follow these choices. An explicit action-specific provider or model stays unchanged when you change the Settings default. If a chosen connection becomes unavailable, reconnect it or choose another; it is not silently replaced.

Choose a provider and model for an action

Open an AI Prompt action’s gear settings, or create/edit an AI Prompt custom action. Both places offer the AI provider and AI model selectors. Keep the inherited default or choose a connected provider and a model for that action. Review action outputs and corrections.

Use Ollama Cloud

Choose Ollama Cloud in Configure AI providers, connect your own credentials and choose an available cloud model. Cloud requests leave your Mac and require Allow online actions. For an on-device model, choose Ollama Local instead.

Connect an OpenAI-compatible API

  1. In Configure AI providers choose API (OpenAI compatible).
  2. Enter your service’s API base URL, including its version/path prefix when required, rather than the full chat-completions endpoint.
  3. Supply an API key if required. A server without authentication can use an empty key. Enter a model ID if model discovery is unavailable.
  4. Connect, choose a model and test a short AI action. Model discovery alone does not confirm generation access.

The server must support OpenAI-style chat completions. Requests go directly to the configured server. This provider requires Allow online actions even for a local endpoint; use only a server you trust.

Privacy and key storage

API keys are stored in the macOS Keychain. Online actions send the selected text and action instructions to the chosen provider when you run the action. Connecting providers does not upload your snippets or clipboard history. Apple Intelligence and loopback Ollama Local run on your Mac.

Troubleshooting

Check Allow online actions, credentials, model access, provider account limits and the configured server address. A disconnected provider may remain referenced by an action; reconnect it or change that action’s selection. Do not share a key or an account authorization code with support.

New provider connections in 2.3.4

In Settings → AI → Configure…, connect DeepSeek, Groq or MiniMax with your own provider key. Each has independent credentials and model choices. Choose an available text model, then set it as the default or select it for an AI action or translation. Online actions require Allow online actions.

Claude Code CLI uses an installed, signed-in Claude Code command-line tool and its existing account access. Choose it in Configure AI providers and follow the local setup/status guidance. Requests run with tools, MCP, hooks, project settings and persistent sessions disabled, with bounded execution. It is an online account connection; it does not make Claude an on-device model.

Existing Apple Intelligence, Ollama Local, Gemini, OpenAI API, ChatGPT / Codex, Claude API, OpenRouter, Perplexity, OpenAI-compatible API, Ollama Cloud and Temporary AI choices remain. Action-specific choices remain when you change the default; an unavailable AI connection is not silently replaced.

Perplexity uses the current Agent connection and Fast, Low, Medium or High effort choices. If a key was revoked, disconnect it and connect the replacement. Keep keys out of screenshots and support messages.

Keep model lists focused

Open an AI model menu and choose Hide model beside a model you do not use. Enable Show hidden models in that menu to reveal it again, then choose Show model. Hiding a model does not delete your provider connection or remove the model from the provider.

Back to all guides