Getting Started

Copair is a terminal-based AI coding agent that works with any model. This guide walks you through installation, provider configuration, and your first session.

Installation

npm (recommended)

npm install -g copair

Homebrew

brew install dugleelabs/tap/copair

npx (no install)

npx copair

Configure a Provider

Copair needs at least one AI provider configured. Create a config file at ~/.copair/config.yaml (global) or .copair/config.yaml (project-level). Choose one provider to start:

Anthropic (Claude)

# ~/.copair/config.yaml
version: 1
default_model: claude-sonnet

providers:
  anthropic:
    api_key: ${ANTHROPIC_API_KEY}
    models:
      claude-sonnet:
        id: claude-sonnet-4-20250514

OpenAI (GPT-5)

# ~/.copair/config.yaml
version: 1
default_model: gpt-5

providers:
  openai:
    api_key: ${OPENAI_API_KEY}
    models:
      gpt-5:
        id: gpt-5

Ollama (Local Models)

No API key needed — just have Ollama running:

# ~/.copair/config.yaml
version: 1
default_model: llama4

providers:
  ollama:
    type: openai-compatible
    base_url: http://localhost:11434/v1
    models:
      llama4:
        id: llama4
        supports_tool_calling: false

Your First Session

Start Copair in any project directory:

cd your-project
copair

You'll see a terminal prompt where you can ask the agent to:

• Read and understand your codebase
• Write new features
• Fix bugs
• Run tests
• Create git commits

Example

Copair will analyze your project structure, create the implementation, write tests, and offer to commit the changes.

While the model is thinking, the spinner shows a rolling preview of the streaming response so you can see progress in real time — useful with slower local models where the wait between tool calls would otherwise look indistinguishable from a hang.

Basic Commands

Next Steps

• Model Switching — Learn how to switch models mid-session
• Git Integration — Automatic branches, commits, and diffs
• Custom Tools — Extend Copair with your own tools