One sentence spawns a full AI consulting department

What it does Agency Orchestrator is a TypeScript CLI that takes a plain-English sentence, assigns roles from a library of 211+ specialist AI agents, and runs them as a DAG workflow to produce a finished report. You can also write zero-code YAML pipelines manually, pass outputs between steps via template variables, and resume from any point if a step needs a redo.

The interesting bit The ao compose command is the standout: the tool uses an LLM to parse your sentence, cast the right characters from its role library, design the dependency graph, and emit a runnable YAML workflow automatically. It also routes requests through CLI tools you might already be paying for—Claude Code, Copilot, Gemini, ChatGPT—so you can skip API keys entirely on several providers.

Key highlights

• 211+ pre-built Chinese roles (and 170+ English) covering product, engineering, finance, marketing, and more, loaded as system prompts.
• Automatic DAG construction and parallel execution based on depends_on; no manual graph wiring needed for composed workflows.
• Resume and partial rerun via --resume and --from, letting you iterate on one step without re-burning tokens for the entire chain.
• Seven no-API-key provider options that piggyback on existing CLI subscriptions or local Ollama, plus standard key-based APIs.
• MCP server mode and integrations for 14 AI coding assistants like Cursor and Claude Code.

Caveats

• The README is CLI-centric; the web Studio and desktop app mentioned in the repository description are not demonstrated in the documentation provided.
• Several “no API key” providers still require existing paid subscriptions (Claude Pro, Copilot, ChatGPT Plus) rather than being truly free.

Verdict Best for developers and tech leads who want structured, multi-perspective AI output without writing Python orchestration. Skip it if you are looking for a chatbot UI or lack access to any supported LLM backend.