AI · Framework
Claude Opus Multi-Agent Orchestration Demo (opus.py)
A Python multi-agent pattern pairing Opus orchestration with Haiku worker instances.
A workflow pattern demonstrating hierarchical multi-agent software development using Claude models. Claude Opus acts as the planner and reviewer, breaking down a high-level goal into sequential technical subtasks that fast, lightweight Claude Haiku instances execute. It serves as an early reference implementation for autonomous coding agents.
The framework
Flow: user enters objective ('create a js html website to draw, like a paint app') into opus.py in the terminal. Opus produces a high-level plan (set up canvas, drawing tools toolbar, color picker/palette, features like brush size/eraser/save/undo-redo, optimize, deploy, maintain). Opus then repeatedly 'Calls Haiku for task:' to implement each subtask — e.g. setting up canvas event listeners (mousedown/mousemove/mouseup drawing logic), color picker and swatches, clear button, then enhancement passes (UX, accessibility/ARIA, code maintainability, testing/docs). Outputs return to Opus for review, and finally 'Calling Opus to summarize task results for objective'. Whole process took 13 minutes 20 seconds.
What it's for
- Build simple web applications from natural language prompts using delegated AI subagents
- Structure hierarchical multi-agent architectures that pair high-capability planners with low-cost executors
- Automate incremental code generation, review, and documentation passes
Key points
- Hierarchical planner-worker architecture dividing tasks between Opus and Haiku
- High-level project decomposition into sequential implementation phases
- Automated task delegation and prompt execution via terminal CLI
- Continuous review and task summarization by the orchestrator model
- Step-by-step modular builds including core logic, UX, ARIA accessibility, and testing
Pricing
Open source. Free script concept; requires user Anthropic API credits.
Details
- Type
- Framework
- Category
- AI
- Topic
- Autonomous Agents
- Open source
- Yes
- Platforms
- Command line
- For
- Developers experimenting with LLM orchestration and autonomous coding agents.
- Alternative to
- Devin, MetaGPT, AutoGPT, ChatDev
- Tags
- claude, anthropic, agents, orchestration, python, automation
Why it was saved
- @skirano: Early demonstration (March 2024) of LLM-orchestrated subagents autonomously building a full app — like Devin — with the full transcript shared.
Compare
Related
Open-source, self-hosted AI agent with persistent memory and messaging integrations.
Open-source, fully local autonomous AI agent that browses and writes code
Open-source platform and framework for building autonomous AI agents
Open-source long-horizon agent harness with Docker sandboxes, memory, and sub-agents
All-in-one AI workspace with autonomous agent tools for content, research, and analysis.
Open vision-to-action foundation model for generalist autonomous game playing
Updated