Overview

Agent Dev Lab is a TypeScript-first toolkit for authoring AI agents and workflows: a headless core (@agent-dev-lab/core), an inspection UI (@agent-dev-lab/web), and a CLI (adl).

What you get

• Agents and workflows as plain TypeScript on top of the Vercel AI SDK (streamText, tool, ModelMessage)
• Persisted run events for waterfalls, SSE tails, and replay (WorkflowStore / SQLite)
• adl init / adl workflow run / adl agent run / adl dashboard for scaffolding, CLI execution, and inspection
• Published adl dashboard is a Nitro serve — restart after registry or .env* edits

Principles

• Runtime/UI split — workflows run from scripts, tests, or server; UI reads persisted output.
• TypeScript-first — plain TS orchestration, no workflow graph DSL.
• AI SDK native — ModelMessage, streamText, tool() without parallel abstractions.
• Colocated prompts — markdown beside code; templates via Handlebars + Zod (createTemplate).

Documentation map

Guides

• Project Setup — the recommended way to start a project (adl init)
• Manual Setup — adding ADL to an existing project by hand
• Inspection UI — adl dashboard, waterfalls, agent conversations, event log
• Gotchas — sharp edges worth knowing about before they surprise you
• Runtime — createAdlRuntime, workflow context, OpenTelemetry
• Agents — adl.createAgent, stopWhen, memory, adl agent run
• Workflows — adl.createWorkflow, ctx.emit, createWorkflowFromAgent
• Project Config — registry, loadAdlProject

API reference

Generated from @agent-dev-lab/core’s own code comments — covers focused APIs that aren’t duplicated as guide pages:

• Package Overview — AI SDK compatibility summary (ModelMessage, OpenTelemetry)
• MessageStore, WorkflowStore, RunEvent (runSeq)
• Template, createTemplate
• createWorkflowFromAgent, createToolFromAgent
• WorkflowObserver, AgentObserver

Use the Core API sidebar for the full export list.

Packages

Package	Role
@agent-dev-lab/core	Headless runtime
@agent-dev-lab/web	Inspection UI
@agent-dev-lab/cli	adl CLI