Description
Fontaine is a versatile terminal AI coding agent engineered to integrate seamlessly with any AI model you choose to configure. This documentation provides a detailed overview of its current capabilities, with each page dedicated to a specific topic, guiding users from initial setup to advanced functionalities.
Read More
The core of Fontaine's design includes a robust guardian system that intelligently judges potentially risky tool calls, enhancing operational safety and reliability. For developers requiring continuous operation, Fontaine offers an AFK (Away From Keyboard) mode, allowing the agent to run unattended while providing recovery guidance for denied prompts and a comprehensive report of all attempted actions.
Installation is streamlined with a one-line installer for Linux and macOS, alongside options for manual archive and source installs, and easy updates via `fontaine update`. The architecture is clearly defined, detailing the workspace layout and the responsibilities of each component. Configuration is highly flexible, covering every settings key, merge rules, workspace trust, model accounts, file locations, and command-line flags.
Fontaine supports various providers, with side-by-side comparisons and minimal examples for each account kind. It includes a suite of built-in tools, categorized by risk class, and offers granular permission settings with deterministic floors and allow rules. Advanced features include subagents for asynchronous tasks, task boards, and user hooks for custom event handling.
For collaborative and complex projects, Fontaine supports MCP servers for declaring stdio and streamable-HTTP servers, instruction files for guiding agent behavior, and a `fontaine doctor` command for comprehensive system checks. Auto memory provides a per-project notes store, gated for writes and auditable. The platform also supports skills, agent files, output styles, and plugins from git-hosted catalogs, managed via the `fontaine plugin` CLI.
The terminal UI is rich with features, offering various commands for managing models, permissions, memory, and sessions, including `/recap` for session reports and `/cost` for token usage. Headless runs are supported for CI/CD environments, and Claude subscription integration is available for specific plans. Telemetry can be enabled for export, with privacy defaults in place. The documentation also covers verifying releases, checkpoints, and terminal setup, ensuring a complete development experience.
- CategoryAI Coding Assistants
- DeploymentTerminal
- PlatformsLinux, macOS
Fontaine's Core Features
Terminal AI Coding Agent
Configurable with Any AI Model
Guardian for Risky Tool Calls
AFK Mode for Unattended Operation
One-Line Installer for Linux/macOS
Detailed Architecture Documentation
Flexible Configuration Options
Support for Multiple AI Providers
Built-in Tools with Risk Classes
Granular Permission Management
Asynchronous Subagents
User Hooks for Custom Events
MCP Server Support
Instruction Files (FONTAINE.md, CLAUDE.md, AGENTS.md)
fontaine doctor for System Diagnostics
Auto Memory for Project Notes
Skills, Agent Files, and Output Styles
Plugin Management via CLI
Rich Terminal User Interface
Headless Runs for CI/CD
Claude Subscription Integration
Telemetry Export with Privacy Defaults
Checkpoints and /rewind for Code Restoration
How to use Fontaine?
Install Fontaine: Use the one-line installer for Linux and macOS, or opt for manual archive/source installs.
Configure Settings: Adjust every settings key, model accounts, and file locations to customize your agent's behavior.
Select Providers: Choose and configure your preferred AI model providers based on your project requirements.
Define Permissions: Set up permission modes, allow rules, and risk classes for tools to control agent actions.
Utilize AFK Mode: Enable AFK mode for unattended operations, receiving recovery guidance and session reports.
Manage Subagents: Create and manage asynchronous subagents and task boards for complex, multi-part tasks.
Extend with Plugins: Install and manage plugins from git-hosted catalogs using the `fontaine plugin` CLI.
Monitor and Debug: Use `fontaine doctor` for system checks and terminal commands like `/cost` and `/recap` for monitoring.
Fontaine's Use Cases
- Automated Code Generation
- Secure Tool Execution
- Unattended Development Tasks
- Custom AI Agent Workflows
- Project Configuration Management
- Multi-Provider AI Integration
- Codebase Auditing and Debugging
- Version Control and Rollback
- CI/CD Integration
- Extensible Development Environment







