Agav Documentation
Agav is a terminal-native AI coding assistant for real repositories. It can inspect code, edit files, run commands, and help verify changes from a CLI-first workflow.
New to Agav? Follow the four-page path below. You do not need to understand agents, tools, skills, or model settings before you begin.
#Get started
- Install Agav
- Connect a model provider
- Complete your first repository task
- Add instructions for your project
The Getting Started overview lists what you need and what you will learn at each step.
#Find what you need
| I want to… | Go to |
|---|---|
| Give Agav a specific file, document, or image | Files and context |
| Continue a previous conversation | Sessions and memory |
| Structure or redirect a larger task | Planning and steering |
| Run Agav from a script or CI job | Non-interactive and CI |
| Automate repeated work | Automation |
| Look up a command or setting | Reference |
| Diagnose a problem | Troubleshooting |
#Learn advanced workflows
After you are comfortable with the basics, use the guides for complete recipes involving subagents, skills, MCP, plugins, and long-running work. Use the features section when you want to understand an individual capability.
Workspace boundary: Agav treats the directory where you start it as the active workspace. Start at your repository root so it can see the whole project.