> ## Documentation Index
> Fetch the complete documentation index at: https://help.partgenie.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# KiCad library generation

> Generate a verified KiCad symbol, footprint, and STEP 3D for an MPN, then download the library (Beta)

**KiCad library generation (Beta)** builds a part library from PartGenie's component database and the datasheet: **Symbol**, **Footprint**, and **STEP 3D**, plus an evidence summary you can check before you drop it into a board.

This is different from downloading a catalog model on the part page. Catalog assets live on the component **CAD Models** tab — see [ECAD & footprints](/en-US/ecad-models). Use generation when you need a KiCad-native library with pin, pad, and visual checks attached.

## Generate a library

<Steps>
  <Step title="Start from Search">
    Open [app.partgenie.ai](https://app.partgenie.ai) and stay on **Search** (not Work).
  </Step>

  <Step title="Ask for a verified library">
    Use the MPN and package in the prompt:

    ```
    generate verified KiCAD for RC0603FR-071KL
    ```

    You can also say *symbol, footprint, and 3D model* for a specific MPN such as `STM32F103RET6` or `OPA810IDBVR`.
  </Step>

  <Step title="Wait for the job">
    Generation can take a few minutes. PartGenie tells you it has started and notifies you **in the app and by email** when the library is ready. You can leave the chat and come back from **Recents** or **⌘K**.
  </Step>

  <Step title="Review Symbol, Footprint, and 3D">
    The finished card shows three previews:

    * **Symbol** — schematic symbol and pin numbers
    * **Footprint** — pad layout (for example 0603 pads 1 and 2)
    * **3D Model** — STEP body

    Status chips on a finished library include **KiCad native import**, **Pin mapping**, and **Visual audit**.
  </Step>

  <Step title="Open the evidence summary">
    Expand **Evidence summary**. Each row is a field (pin count, pitch, body size, pad size), the value used, the **datasheet page** it came from, and a status:

    * **Calculated** — derived from datasheet or package data
    * **Assumed** — a default the engine filled in (pitch is a common one)

    Treat **Assumed** rows as items you must confirm.
  </Step>

  <Step title="Download">
    Click **Download** on the part card and import the files into your KiCad project libraries.
  </Step>
</Steps>

<Frame>
  <img alt="Generated KiCad card with Symbol, Footprint, 3D Model, and Download" src="https://mintcdn.com/auraloomllc/hquMDjp8DWjpUjd3/images/kicad-library-generate.png?fit=max&auto=format&n=hquMDjp8DWjpUjd3&q=85&s=8831dddbe328cb42b2754b068853053b" width="1024" height="682" data-path="images/kicad-library-generate.png" />
</Frame>

<Frame>
  <img alt="Evidence summary table with datasheet page and Calculated or Assumed status" src="https://mintcdn.com/auraloomllc/hquMDjp8DWjpUjd3/images/kicad-library-evidence.png?fit=max&auto=format&n=hquMDjp8DWjpUjd3&q=85&s=00c2eabc9d5280301019f1d255e3b086" width="1024" height="682" data-path="images/kicad-library-evidence.png" />
</Frame>

## Before you fabrication-release

<Warning>
  Libraries are **Beta**. Before you use one, check pitch, pads, pin 1, and the 3D model against the datasheet. An Assumed pitch or a wrong pin-1 mark will not be caught by a visual thumbnail alone.
</Warning>

* Open the cited datasheet page from the evidence table (see [Datasheet citations](/en-US/datasheet-citations))
* Confirm pin 1 on both the symbol and the footprint
* If something is wrong, reply in the same chat with the correction — and send feedback so the generator can improve

## Where files go afterward

Downloaded libraries are yours to import into KiCad. The chat that produced them stays in **Recents**. Related reports and files can also land in [Knowledge](/en-US/knowledge) when PartGenie saves the artifact.

## Related

* [ECAD & footprints](/en-US/ecad-models) — catalog models already on the part
* [Design mode](/en-US/design-mode) — architecture and BOM before you draw the board
* [AI Search](/en-US/ai-search)