The asset model: what Berth actually shows you

Everything in Berth is built on one idea: the files that shape how your AI agent behaves are assets — and assets deserve to be visible, searchable, and related to one another.

An asset is any file or configuration that shapes how an AI agent works. Berth groups them into two families you will see throughout the app.

Instructions — what the agent should do

Memories (CLAUDE.md / AGENTS.md), Skills, Subagents, Commands, Output Modes, and Agent Teams. These are the text that guides behavior before and during work.

Capabilities — what the agent is able to do

MCP servers, Hooks, Permissions, Environment variables, Status lines, and Plugins. These define the agent’s runtime powers and boundaries.

v0.1 is read-only: Berth reads these files to display them and never writes to them. Credentials such as API keys are detected for status only and never shown.