Developer Overview
This section documents the underlying Toolang design. It is meant for people who need the model behind the CLI, runtime, storage layout, and integration boundaries.
What this section covers
The developer docs are organized around stable runtime concepts:
- filesystem layout and agent identity
- capability resolution and scope rules
- execution, scheduling, and durability
- tasks, chores, and collaboration
- local APIs and shared projections
- plugins, memory, and implementation constraints
Each page owns one topic so the same rule does not need to be repeated in multiple places.
Core vocabulary
toolang rootthe local Toolang system directoryagent homethe local directory that hosts one or more.toofilesagent roomthe private machine-managed area for one agentagent_urithe canonical identity stringagent_ida short stable hash derived fromagent_uriref,inline,localthe three capability formsagent,shared,globalthe three capability scopesruntime loopa long-lived trigger source such asserver,poll,hook, orpulseexecution strategythe strategy used to complete one turnrunone continuous active interval of one agent processthreada durable execution contextturnone complete handling attempt inside a threadstepone internal action inside a turn
Design rules
- local files and databases are the primary execution truth
.toosource is authored state, not the final runtime inputtoolang syncmaterializes exact runtime inputs before execution- runtime loops decide when work starts
- execution strategies decide how one turn completes
- plugins stay at the edge and do not replace runtime lifecycle or scheduling
Read next
- Read Layout and Storage for filesystem layout and canonical identity.
- Read Capability Model for refs, scopes, and sync materialization.
- Read Execution Model for runs, threads, turns, steps, and scheduling.
- Read Task Model for durable task semantics.
- Read Control Surfaces for the CLI, registry, agent API, and bus API.