esp32-agent-control-demo

Chinese Version

This repository is a compact control stack for AI-agent-driven ESP32 interaction.

I use it to think through the practical boundary between model-facing tools and device-facing commands: how an agent should call functions, how a bridge should normalize them, and how an ESP32 runtime should route them into screen, LED, servo, or motion behavior.

• Repository: 78/xiaozhi-esp32
• Description: An MCP-based chatbot
• License observed: MIT

The upstream project is an important reference because it already exposes device-side MCP control concepts. This repository keeps my smaller, easier-to-read version of that control path in the form of schemas, examples, and a runnable translator.

• an MCP-style tool catalog for device interaction
• a bridge script that translates tools/call payloads into compact device commands
• JSON schemas for commands and status reports
• clean-room firmware examples for routing commands on an ESP32 device

• `docs/upstream-reference.md` upstream context and repo boundary
• `docs/architecture.md` control flow from agent to device
• `docs/tool-catalog.md` tool naming and parameter model
• `schemas/device-command.schema.json` command envelope
• `schemas/device-status.schema.json` status envelope
• `examples/mcp` MCP-style requests and responses
• `examples/device` translated device-side examples
• `bridge/mcp_to_device_bridge.py` runnable translator
• `firmware/command_router_example.cpp` command routing sketch
• `firmware/tool_registry_example.cpp` compact tool r