Concepts overview

The short version

Lasso is an agent-native monorepo: one workspace where agents can discover repositories, create isolated worktrees, load shared skills, and preserve durable handoffs.

The monorepo is a workspace and coordination surface, not one forced source tree. External repositories remain independent clones; subdirectory projects can remain true monorepo trees.

project identity → stable checkout → task worktree → agent execution

Read next

• Core mechanism — the lifecycle from lasso init to an isolated task surface.
• Extension model — how modules add conventions, checks, skills, hooks, and scripts.
• Agent runtimes — the Codex baseline and thin follower adapters.

What belongs where

Concern	Owner
Current goal, plan, review, session, memory	Agent runtime
Project identity and stable checkout bootstrap	Lasso
Canonical worktree placement and safe GC	Lasso, with Git as source of truth
Durable facts and exceptional handoffs	Lasso work items
Build, test, deploy, secrets, service health	Owning project or host
Optional conventions, checks, skills, and hooks	Versioned modules

Lasso stays useful by knowing where to stop. It creates a navigable workspace around project toolchains; it does not create a second build system or agent task runtime.

Product vs generated workspace

Lasso	The product (this repository)
Your workspace	A workspace created by lasso init

Change a workspace capability → change Lasso or add a module.

Change your repositories, machines, or business docs → change your workspace.