Open Hardware & RISC-V

Konnect brings AI-assisted PCB design to KiCad 10 via Model Context Protocol

Key Takeaway

A native Rust plugin enables LLMs to manipulate KiCad 10 schematics and boards through 228 tools exposed via the Model Context Protocol, replacing deprecated SWIG bindings with the official IPC API.

Summary

Developer mixelpixx has released Konnect, a native KiCad 10 plugin that bridges large language models to printed circuit board design through the Model Context Protocol (MCP). The tool ships as a single Rust binary of approximately 20-25 MB and exposes 228 specialized functions across 21 toolsets including schematic capture, PCB layout, electrical and design rule checking, design review, JLCPCB part lookup, and manufacturing export. Unlike its predecessor KiCAD-MCP-Server, which relied on Python, TypeScript, and SWIG bindings that KiCad is phasing out, Konnect communicates with KiCad 10 through the official IPC API using protobuf over NNG, ensuring that AI-generated edits participate in the application's native undo and redo system.

The plugin employs a demand-loaded toolset architecture that reduces initial context from roughly 23,000 tokens to about 2,000 tokens, loading additional capabilities only when needed. Schematic modifications are handled by an S-expression engine that performs atomic writes while preserving UUIDs to maintain design integrity. A standalone schematic viewer with live re-rendering is also included for monitoring changes as they occur.

Architecture and integration with KiCad 10

  • Rust binary delivery: Konnect distributes as a single compiled binary of 20-25 MB installed from a ~9 MB download, eliminating dependency management and version matching requirements for users.
  • MCP toolset routing: The plugin registers 228 tools across 21 categories but exposes only a ~2K token starter set initially, fetching additional toolsets on demand rather than consuming ~23K tokens in every context window.
  • Official IPC API replacement: The rewrite abandons SWIG bindings in favor of KiCad 10's native IPC mechanism, using protobuf serialization over NNG (nanomsg-next-generation) for inter-process communication.
  • Undo/redo integration: Because edits flow through KiCad's official API rather than direct file manipulation, AI-driven changes integrate with the application's built-in undo and redo stack.
  • S-expression schematic engine: Schematic file modifications use an S-expression parser with atomic write operations and UUID preservation, paired with a standalone viewer that re-renders schematics as underlying files change.

Why it matters

Konnect represents a meaningful step toward practical AI assistance in open hardware workflows by addressing the fragility of earlier MCP integrations that relied on deprecated bindings. For hardware designers using KiCad, the demand-loaded architecture makes LLM-assisted design economically viable with current token pricing, while the Rust implementation and official API usage suggest maintainability as KiCad evolves. The plugin's approach may serve as a reference for how other complex desktop applications can expose granular, safe AI manipulation without sacrificing user control or application integrity.

Original reference: blog.adafruit.com →