Automation and agents

Edit animated scenes in a Raylight browser project through your assistant. It can read a scene, change it and show rendered frames for review.

Creating a team interface for a specific program.

Distribution of multiple agents by tasks and budget.

Standalone agent with memory and tools.

Build one n8n workflow: turn a practice request into a clear summary. Start with manual runs and no external sending.

Build an MCP server that reads practice notes from one folder. The SDK helps write the server; connecting it to your Codex is a separate check.

Python tool for voice dialogs. Codex can help write the app; it is not a ready phone line. Separately needed are model, speech, connection, script, and delay check.

Python tool for task and role chains. Do not count each agent as independent fact confirmation. Check input, output, repeats, and errors of each step.

Separate variants: library, CLI, and cloud. README describes usage with Codex, but actual connection in our research is unverified. Start with a page reading example without

Anthropic's educational example for Claude Managed Agents. README clearly warns the project is unsupported, and the explanatory process consumes more tokens than specialized

Voice web interface with Claude Agent SDK and local bridge. README requires Claude Code with subscription, Node 20+, and Chrome/Edge. ElevenLabs is optional; a fallback uses