Local and CLI Providers

Local and CLI providers allow Kapinote to run inference through software already installed on your Mac. Currently supported integrations include Ollama, Gemini CLI, Claude Code, GitHub Copilot CLI, and OpenAI Codex CLI.

1. Local Ollama

Local Ollama does not require an API key, but you must install Ollama, download a model that fits the Mac's unified memory, and keep the service running. Unlike the developer CLI tools below, Ollama is the most approachable local option for someone without a cloud account or API key.

Follow the dedicated complete Ollama setup guide. It includes:

• Beginner-friendly installation, with the model downloaded from inside Kapinote
• Qwen recommendations for M1 through M5 Macs with 8 GB to 128 GB of memory
• Exact values for every Kapinote configuration field
• Help for command-not-found, connection, missing-model, and memory errors

2. CLI providers integration model

Connecting a CLI provider requires zero API key input in Kapinote. Kapinote invokes the locally authenticated binary via a non-interactive subprocess:

1. Install the official command-line tool.
2. Complete the tool's native login/authentication in your macOS Terminal.
3. Execute a basic test query in Terminal to ensure the account session is active.
4. In Kapinote under Settings → AI → CLI, select the provider and click Verify.
5. Model Aliases: Leave Model blank to inherit the CLI's default, or specify a valid alias (e.g., sonnet, opus, gemini-2.5-flash).

Environment Note: Kapinote must locate the binary within your system PATH. If verification indicates the command is missing, completely quit and restart Kapinote. See Troubleshooting.

3. Tool-specific guides

Gemini CLI

• Follow the Gemini CLI Getting Started and Authentication Guide.
• Run gemini --version in Terminal to confirm availability. Usage inherits your active Google account credentials.

Claude Code

• Install and authenticate Claude Code per the Anthropic Setup Guide.
• Verify claude responds in Terminal. Leave Model blank or specify aliases like sonnet or opus.

GitHub Copilot CLI

• See Set up Copilot CLI.
• Install command: npm install -g @github/copilot.
• Run copilot login to authorize your GitHub Copilot subscription.

OpenAI Codex CLI

• Install via npm install --global @openai/codex and execute codex login.
• Check status using codex login status. See the Codex CLI documentation.

Privacy and execution lifecycle

• Ollama: Installation and model downloads require internet access. When using a local model tag through localhost, meeting text is processed on the Mac. See Ollama privacy boundaries.
• CLI Tools: The process executes locally on macOS, but transmits transcript context to the provider's cloud according to that tool's native account policy and data agreements. Kapinote operates as an orchestrator and does not intercept or modify each tool's underlying telemetry or billing.