Agents In Use
What’s new todayLast update
1005 additions

How to use Claude to check a KiCad converter PCB

Claude Code designed a KiCad buck-converter PCB through MCP. Datasheet, netlist and routing checks corrected errors; no fabricated load-tested supply was shown.

Published jlcpcb.comAgent: Claude
OUR ADAPTATION

Suggested prompt

Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.

Review and build this KiCad buck-converter design against the component datasheets. Verify pin-by-pin netlist connectivity, schematic-to-PCB parity, courtyards, clearances and final DRC. Report unresolved thermal and mechanical checks and stop before fabrication or claims about electrical load performance.

What happened

The buck-converter design reached clean connectivity, parity and DRC checks after multiple corrections. The source does not report fabrication or electrical load tests.

Reported by JLCPCB — Engineering blog

How the agent did it

  1. The operator reviewed the reference circuit against component datasheets before accepting the schematic.
  2. Claude built and corrected schematic nets after pin-by-pin review exposed label and placement problems.
  3. The layout was enlarged and rerouted until connectivity, board-edge and parity checks passed.

Where a person stepped in

The operator checked the datasheet, reviewed the schematic pin by pin and verified layout constraints and remaining silkscreen warnings.

JLCPCB tested Claude Code with KiCad MCP by designing an XL4015 constant-current, constant-voltage buck converter. The reference circuit was a starting point rather than a guarantee: the operator's datasheet review caught incorrect control-pin wiring and a regulator that could not accept the intended maximum input voltage.

Claude built the schematic through tool calls, but multi-unit component placement and label collisions introduced further problems. A pin-by-pin netlist review found a connection mistake that appeared only as an ERC warning. The operator and agent corrected the intended nets instead of accepting the absence of severe schematic errors.

The original small board outline also could not accommodate the component courtyards. The layout grew, placement checks addressed overlaps, and routing had to be corrected where a feedback trace crossed an input connection. A mismatch between router clearance and board-edge rules required another routing pass.

The final design reported no DRC errors, unconnected items or schematic parity errors. It was not a fabricated, load-tested power supply. This account's useful result is the checked design artifact and the sequence of corrections that produced it. Human datasheet, netlist and geometry checks remained necessary even when the tool-generated schematic looked orderly.

Prompt
Suggested prompt (written by Agents In Use, not the author's)

Source: Engineering blog · JLCPCB

Frequently asked questions

What did datasheet review catch?

Incorrect control-pin wiring and a regulator unsuitable for the intended input voltage.

Why inspect the netlist?

A label collision caused a wrong connection that did not appear as a severe ERC error.

Was a working supply demonstrated?

No. The source reports a checked PCB design rather than fabricated load-test results.

More details

About this account

The operator reviewed the reference circuit against component datasheets before accepting the schematic. Claude built and corrected schematic nets after pin-by-pin review exposed label and placement problems.

Primary HTML successfully retrieved and read despite web-fetch failure. Design passed final DRC but no fabricated or load-tested converter is reported. Company-published firsthand case study. Suggested prompt written by Agents In Use; no original task prompt was published in the verified account.

Credit: JLCPCB — Engineering blog

Source published 14 Oct 2026

Source checked 14 Oct 2026

Tools and topics

Claude Code · KiCad 10 · KiCad MCP · XL4015

Power electronics · Netlist verification

Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.