Guides

These guides cover the user-visible surface of Egghead: the CLI, the TUI, the web interface, and the MCP server. They assume you have a working shell, an editor, and an API key for at least one LLM provider.

If you have already used CrewAI, Paperclip, or OpenClaw, a few of Egghead’s design choices will look unfamiliar at first, and that is on purpose. Records live as plain Markdown files in a directory you choose, not inside an internal database. Agents are defined by those Markdown files, not by Python classes or YAML roles. Capabilities are written into an agent’s frontmatter ahead of time, not approved through “Allow once” prompts at runtime. Most chat in Egghead happens in a shared transcript that every participant reads, not through a dispatcher that routes each turn. The rest of the documentation explains how each of those choices works in practice and what it costs you when it doesn’t fit.

If you are new to the system, read Getting started first; it gets you to a running node with one agent in about ten minutes. After that, the order in the left rail follows the order things become useful: records and record classes describe what you store; the agents section describes what runs against that store; the capabilities section describes what those agents are allowed to do; and the rest of the rail covers integration, day-to-day operations, and evaluation.

If you want the why before (or after) the how, the founding essay is an optional long-read on what Egghead is for and why it is shaped the way it is.

Every guide is written against the current release. If a CLI flag, configuration key, or function signature in these pages does not match the binary you are running, the binary is right and the guide is stale; please file an issue.