MakerLord is an AI assistant for people building electronics. It covers the maker’s journey from idea to simulation, prototype and product. It’s live at makerlord.dev, and joining needs an invite.
Most AI assistants will happily tell you to wire an LED straight to 5 V. In MakerLord the model proposes and explains, and deterministic rules decide. Gates like “open the power gate” or “generate firmware” are enforced by the engine, not by asking the model nicely. There is no dismiss_finding tool. You cannot call what was never defined.
The journey has seventeen stages in four acts. The first two are built. Design covers feasibility research with cited evidence, requirements with units, and a block architecture with power budgets. Prove covers simulation on a virtual bench with ngspice, a real breadboard build in safety-ordered steps, firmware derived from the wiring, and guided debugging that proposes one measurement at a time. The schematic, breadboard view, netlist and firmware are all projections of one circuit model.
You can drive it three ways: your own API key for almost any model provider, an agent you already run (Claude Code, Codex, Gemini CLI or another ACP agent) through a local bridge, or MCP. All three use the same 50-tool registry, so the gates hold whichever model is in charge. Parts come from the Fritzing library with electrical limits written by hand, and are never invented. Two test suites, a circuit danger corpus and a firmware danger corpus, block any release that fails them.
I built it at the end of July 2026. It went from spec to a deployed app in four days. Agents wrote most of the code, and the decision log, now over fifty entries, is how I kept the decisions mine (Working with a team of agents). The last two acts, from PCB to production, aren’t built yet. The code is MIT-licensed on GitHub.