USB to Serial Adapter for Mac: macOS Matches 106 IDs

Search for a CH340 or PL2303 driver for a Mac and you get two kinds of advice: install the chip vendor's driver, or don't, because macOS already has one. Both are right, for different cables. Which one applies to the adapter in your cart depends on a USB vendor and product ID, and on whether that pair is in a table Apple ships.

So I read the table. The Mac mini that runs this business (Mac16,10, M4) is on macOS 26.4.1, build 25E253. I loaded the Info.plist of every USB serial driver under /System/Library/DriverExtensions and collected the IDs each one matches. Then I compared them with the ID tables in Linux's usb-serial drivers for the same four chip families. I don't own a USB serial adapter. ls /dev/cu.* on this machine lists Bluetooth-Incoming-Port and debug-console, both built in. Everything I say about specific cables below comes from the vendor, the listing, or owner reports.

What macOS 26.4.1 ships

Five DriverKit extensions match serial adapters by ID. Together they hold 106 unique vendor:product pairs:

Driver (.dext)	Personalities	Unique IDs	What they cover
AppleUSBFTDI	97	96	FTDI FT232R, FT2232, FT4232H, FT232H, FT-X (FT230X/FT231X), plus dozens of radio programming cables and lab instruments that use FTDI chips under their own vendor IDs
AppleUSBPLCOM	5	5	Prolific 067b:2303, 2304, a100, e1f1, and ATEN's 0557:2008
AppleUSBCHCOM	4	2	WCH 1a86:7523 (CH340/CH341) and 1a86:55d4 (CH9102F)
AppleUSBSLCOM	3	2	Silicon Labs 10c4:ea60 (CP2102, CP2102N, CP2104) and ea70 (CP2105, both ports)
AppleUSBSerial	1	1	One Google device, 18d1:503b

There is a sixth path that needs no ID at all. AppleUSBCDC.kext 5.0.0 matches by USB class: communications-class devices, composite devices with an interface association, and class-0 devices that declare their own interfaces. Anything that speaks standard CDC-ACM, such as a Raspberry Pi Pico, most Arduino boards with native USB, or Raspberry Pi's Debug Probe, falls into it and shows up as /dev/cu.usbmodem*.

For comparison, the Linux master branch, fetched the same morning, lists 874 unique IDs in ftdi_sio.c, 204 in cp210x.c, 78 in pl2303.c and 6 in ch341.c: 1,162 against Apple's 106. Most of that gap is OEM products that reuse a vendor's chip under their own ID, which few people will ever plug into a Mac. The gap that matters for a buyer is smaller and more specific.

The 19 chip IDs macOS doesn't match

I took the IDs each chip vendor assigns by default, which are the ones a generic adapter carries, and checked all 30 against Apple's plists:

• Prolific's G series is missing entirely. Prolific's PL2303GS page lists product IDs 23A3 through 23F3 alongside 2303 and 2304. Apple matches the last two only. A cable built on a PL2303GC, GS, GT, GL, GE or GB enumerates on the USB bus and gets no serial port until you install Prolific's driver.
• FTDI's HP series is missing. 6040–6045 are the FT2233HP, FT4233HP, FT2232HP, FT4232HP, FT233HP and FT232HP, FTDI's chips with USB-C power delivery built in. 6048, the automotive FT4232HA, is missing too. The chip that nearly every FTDI console cable uses, the FT232R at 6001, is covered.
• WCH: two of the four. CH340 at 7523 and the CH9102F at 55d4 match. 5523 (CH341A) doesn't, and neither does 7522, a second CH340 ID that Linux added in June 2020. The CH343 (55d3) is a CDC-class device, which Linux drives with cdc-acm, so on a Mac it would go to AppleUSBCDC. I can't confirm that without the chip.
• Silicon Labs: the four-port CP2108 (ea71) is not matched, and neither are the alternate IDs ea63, ea7a and ea7b, which Linux labels "Windows Update".

One thing the plists don't check: none of the 106 personalities matches on bcdDevice, the chip revision field. Any cable that reports 067b:2303 matches AppleUSBPLCOM, whether the chip inside is a current PL2303HXD, an end-of-life HXA, or a clone. What the driver code does after it attaches, I can't tell. The PLCOM binary has almost no readable strings. The FTDI one has a single fallback message: "Invalid chip ID (%x), assuming F1232AM".

To see which case a cable you already own falls into, plug it in and run:

# vendor and product ID of everything on the USB bus
ioreg -p IOUSB -l -w0 | grep -E '"USB Product Name"|"idVendor"|"idProduct"'

# did a serial driver attach?
ls /dev/cu.usbserial* /dev/cu.usbmodem* /dev/cu.wchusbserial* 2>/dev/null

# unique IDs each built-in serial driver matches
python3 -c '
import plistlib, glob
for p in sorted(glob.glob("/System/Library/DriverExtensions/com.apple.DriverKit-AppleUSB*.dext/Info.plist")):
    ids = {(v["idVendor"], pid) for v in plistlib.load(open(p, "rb")).get("IOKitPersonalities", {}).values() if "idVendor" in v
           for pid in v.get("idProductArray", [v.get("idProduct")])}
    if ids: print(p.split("DriverKit-")[1].split(".dext")[0], len(ids))
'
# AppleUSBCHCOM 2 · AppleUSBFTDI 96 · AppleUSBPLCOM 5 · AppleUSBSLCOM 2 · AppleUSBSerial 1

ioreg prints IDs in decimal: 1659 is Prolific's 0x067b, 1027 is FTDI's 0x0403, 6790 is WCH's 0x1a86, 4292 is Silicon Labs' 0x10c4. The last command prints the counts in the table above; it expands idProductArray, which a plain grep of the plist would miss.

When the vendor driver is the problem

Apple began shipping AppleUSBCHCOM and AppleUSBPLCOM with macOS 10.14 Mojave, according to Decisive Tactics, who make the Serial terminal app. The same note calls Prolific's older kext "known to be unstable and can cause kernel panics and data loss" and recommends removing it. The negative reports I found, from owners rather than vendors:

• In 2015, a Hacker News commenter described WCH's own Mac driver as unsigned for years and causing kernel panics on the OS X release of the day.
• Built-in drivers have their own bugs. One owner found Apple's FTDI driver ignoring hardware flow control with one device, and it kept sending after the other end signaled not ready. FTDI's own driver worked.
• Another had to install Prolific's driver despite the bundled one and blamed clones that ship under Prolific's official IDs. A third threw out a set of TTGO display boards whose USB serial chip never worked on a Mac "no matter what drivers", though they worked on a PC.
• An Apple Community thread from February 2023: Prolific's 1.5.1 driver refused to install on macOS 13.2 Ventura. Prolific now ships a DriverKit version as the PL2303 Serial app (version 1.4, March 2025, macOS 11 or later). Its App Store reviews include "the Prolific PL2303 driver did not load correctly for me", a device that never appeared under /dev, and hard crashes while flashing a Pi Pico.

WCH's current driver, ch34xser_macos, covers CH340 through CH9143 and names ports tty.wchusbserial*. Its README says nothing about Apple's built-in driver. If your CH340 already shows up as /dev/cu.usbserial-*, AppleUSBCHCOM has it, and you don't need a second driver.

Update 2026-10-25: USB OBD2 car diagnostic cables use the same CH340, FT232R and PL2303 chips. Apple’s tables still match 106 IDs on 26.4.1, but iPhone apps running on a Mac can’t open serial ports at all. The OBD2 scanner for Mac census covers which apps can use a USB cable.

Update, 2026-10-25: the same match tables cover Meshtastic LoRa radios. The CP2102 on a Heltec V3 and the CH9102F on a T-Beam v1.2 hit the two rows above, and nRF52840 boards fall through to Apple's class-matched CDC-ACM driver. My Meshtastic on Mac check has the board table and why Meshtastic's own docs still send you to the vendor drivers.

What to buy for a Mac mini M4

The M4 Mac mini has no USB-A ports: two USB-C on the front, three Thunderbolt 4 on the back. A USB-C cable avoids an adapter dongle and one more point of failure. Everything below is from the listing and the vendor. I haven't used any of them.

As an Amazon Associate I earn from qualifying purchases. Prices and stock were read with a New York delivery address on 2026-09-19.

• Switch, router or firewall console (RJ45): the DSD TECH SH-RJ45B, USB-C to RJ45 with an FT232RL according to the listing. $14.99, in stock, 4.5 stars from 168 ratings. FT232R is 0403:6001, in Apple's table.
• DB9 RS-232 gear (UPS, PDU, older lab kit): the OIKWAN USB-C to RS232 DB9, FT232RL per its bullets. $13.99, in stock, 4.5 stars from 464 ratings. Its bullets still tell you to download FTDI's driver first. Plug it in and check for /dev/cu.usbserial-* before you do.
• Raspberry Pi 5 boot console: the Pi 5's bootloader prints to UART10 on a small three-pin debug connector at 115200 8N1, per Raspberry Pi's docs. The Raspberry Pi Debug Probe has the matching JST-SH cable, and its firmware declares a standard CDC-ACM interface under 2e8a:000c. That is the class-matched path, so no ID table is involved. The Amazon listing didn't show a buy-box price when I checked, so I'm linking the maker's page.

Listing chip claims are claims. Clones of the FT232R and PL2303 carry the genuine IDs, which is the problem the Hacker News comment above complains about. After it arrives, run the ioreg line above. If the IDs match the table but no /dev/cu.* appears, the chip inside may not be what the box says. That's a return, not a driver hunt.

This is the same kind of check I ran for 10GbE adapters for Mac (macOS matches 14 of 25 Aquantia IDs), for 2.5GbE USB adapters on a Mac mini, and for the USB storage quirk tables behind 3.5" HDD enclosures. If the console cable is for a small rack, the 10-inch rack switch survey lists which models are managed. On a headless Mac, setting up without a monitor goes over the remote side.

Update 2026-09-19: Bluetooth dongles are decided somewhere else. There is no ID table in a plist; bluetoothd carries it in code, and I read it out in which USB Bluetooth adapter works on a Mac.

Update 2026-10-01: barcode scanners have a serial mode too. Of 19 scanner manuals I read for the barcode scanner for Mac post, only Honeywell's say their virtual COM mode uses macOS's built-in CDC class driver; Zebra ships a Windows-only CDC driver. In keyboard mode a scanner needs no driver, but it types through the Mac's keyboard layout.

Update 2026-10-05: 3D printers that take jobs over USB hit the same table. In whether a 3D printer is Mac compatible I read six slicer installers; all run natively on Apple silicon, and the USB path comes down to the printer's serial chip.

Update 2026-10-10: The same chip-default IDs show up in VoiceOver’s braille drivers. Four of the 27 bundles in macOS 26.4.1 claim 0403:6001 and two claim 10c4:ea60, because many older displays sit behind a stock FTDI or CP210x bridge. The full rule census is in braille display for Mac.

Update 2026-10-10: RTL-SDR radio dongles are the opposite case. None of macOS 26.4.1’s 987 driver plists matches their RTL2832U IDs, so libusb reaches them with no driver at all. The app bundles are where they break, covered in RTL-SDR on Mac. Ham radio rigs and digital-mode interfaces that use the Silicon Labs chip, and the apps that drive them over these ports, are covered in ham radio software for Mac.

Update 2026-10-11: Zigbee coordinators are mostly these same bridges. Of the 19 sticks in Home Assistant’s USB discovery list, 14 match an Apple driver by ID, and the failures Mac owners report happen when the stick is handed to Docker or a VM. Details in Zigbee dongle on a Mac.

Update 2026-10-11 (GPS): The Prolific G-series gap above now catches a popular GPS receiver. GlobalSat’s BU-353N and BU-353N5 use a PL2303GC (067b:23a3), so they need Prolific’s driver on a Mac, while u-blox receivers match Apple’s CDC-ACM driver. See GPS receiver for Mac.

Related: the usual way to open the port is screen /dev/cu.usbserial-XXXX 115200 with the screen already in macOS. Which version that is, and what it can't do, is in screen command on Mac.

Related: budget vinyl plotters talk over these same FTDI and CH340 chips, and Sure Cuts A Lot 6.085 still checks for drivers by old kernel-extension IDs. What that means before you install a driver is in vinyl cutter for Mac.

Update 2026-10-21: A board that enumerates fine can still fail to program. In Arduino IDE on Apple silicon, the compilers and uploaders for all 19 of Arduino's own board cores include Intel-only tools, so the port shows up and the upload stops with "bad CPU type in executable". Which boards build without Rosetta is in Arduino IDE on an Apple silicon Mac.

Update 2026-10-22: The CH341A has a second USB ID that this table doesn't cover. With its jumper set for programming it enumerates as 1a86:5512, which no macOS driver claims, and flashrom, IMSProg and ch341eeprom drive it through libusb with no driver at all. Which tool handles which chip family is in CH341A programmer on Mac.

Update 2026-10-22: Drone flight controllers fall under the same rules. STM32 and AT32 boards enumerate as class-matched CDC devices, so Betaflight on Mac needs no driver; a USB-C to USB-C cable is the more likely culprit when no port appears.

Related: Davis weather station USB loggers are a common reason to need a Silicon Labs serial driver on a Mac, and one owner, along with WeatherCat's documentation, says it needs updating with each major macOS release. What works instead is in weather station for Mac.

Update 2026-10-26: the Flipper Zero is another native-USB CDC device that needs none of these bridge drivers. The details are in Flipper Zero on Mac.

Update 2026-10-27: if the adapter is for flashing ESP32 boards, the driver is rarely the problem. Ten of 13 official Espressif devkits use a CP2102N that macOS matches out of the box. What goes wrong more often is the tool: pip on a stock Mac installs an old version. I tested four install paths in esptool on Mac.

Method: Apple's IDs come from my Mac mini (Mac16,10) on macOS 26.4.1 (25E253), read 2026-09-19 with Python's plistlib from the Info.plist of AppleUSBFTDI, AppleUSBPLCOM, AppleUSBCHCOM, AppleUSBSLCOM and AppleUSBSerial under /System/Library/DriverExtensions, plus the class-based personalities of AppleUSBCDC.kext. Only matching was read. Whether a driver accepts a given chip revision after matching is not something a plist shows. The Linux counts are unique vendor:product pairs in the id tables of ftdi_sio.c, cp210x.c, pl2303.c and ch341.c on the master branch, fetched the same day. Owner reports are four Hacker News comments read through the Algolia API, one Apple Community thread and the App Store reviews of Prolific's app. Amazon prices, stock and ratings were read with a New York delivery address on 2026-09-19. I have no USB serial adapter, so nothing here was tested with hardware.