Introduction

Braide is a desktop-style web application for managing and interacting with AI agents using the Agent Client Protocol (ACP). It provides a rich interface for creating projects, running agent sessions, managing agent processes, and integrating with GitHub.

What You Can Do

• Create projects linked to directories on your local machine
• Run agent sessions with real-time streaming of agent responses, tool calls, and permission requests
• Manage agents — install, start, stop, and monitor agent processes
• Attach GitHub issues and pull requests to sessions so agents have full context about the work
• Use an embedded terminal alongside your agent sessions
• Configure per-project settings like approved commands for automated tool use

How It Works

Braide connects to agent registries to discover available agents. Once agents are downloaded and running, you can create sessions that send prompts to agents and stream their responses in real time. The entire interaction is displayed as a series of events — chat messages, tool calls, permission requests, and configuration updates.

Key Concepts

Before diving in, it helps to understand a few terms:

• Project — A named reference to a directory on your filesystem. Projects group related sessions together.
• Session — A single conversation with an agent. Sessions track the full history of prompts and responses.
• Agent — An AI agent process that runs locally and responds to prompts via the ACP protocol.
• Registry — A server that lists available agents for download and use.

Next Steps

• Install Braide to get set up
• Follow the Quick Start to run your first session
• Read about core concepts in more detail