Target support

Through both built-in support and CMSIS-Packs, pyOCD supports nearly every Cortex-M MCU that is available on the market.

In addition, because pyOCD dynamically inspects the target’s debug infrastructure, basic debug functionality is enabled for any target that correctly implement the CoreSight architecture.

Target types

When pyOCD connects to a device, it needs to know what type of device (target) it is controlling. The type of device is known as the “target type”. The target type defines the flash programming algorithm and determine the device’s memory map and other important information. Each target type is identified by a short string that is either the full or partial part number for the device. For example, “k64f” or “stm32l072”.

Target types are either built-in to pyOCD or come from CMSIS-Packs. The sections below describe more about the sources of target support.

To see the available target types you can run pyocd list --targets. This command will print a table of supported target types, including the target name, vendor, and part number. In addition, it prints whether each target target is built-in or comes from a CMSIS-Pack.

Generic target type

PyOCD furnishes a generic target type named “cortex_m”. This target type will be able to connect to and debug almost any Cortex-M device that correctly implements the CoreSight architecture. However, flash memory cannot be programmed, and a memory map is not provided. In addition, it may not work for certain devices that require special handling of operations such as reset and halt.

Because of the limitations of the generic “cortex_m” target, the warning shown below will be logged if the target type is “cortex_m”.

0000183:WARNING:board:Generic cortex_m target type is selected; is this intentional? You will be
able to debug but not program flash. To set the target type use the '--target' argument or
'target_override' option. Use 'pyocd list --targets' to see available targets.

Setting the target type

There are two ways to specify the target type.

The first is to pass the --target command line argument to the pyocd tool and pass the target type name. This argument must be passed every time you run pyocd with a subcommand that connects to the target.

Another method is to set the target_override session option in a pyocd.yaml configuration file. The configuration file documentation describes how to do this for a specific debug probe instead of globally.

Target type auto-detection

Certain on-board debug probes know the type of target with which they are connected. The Arm DAPLink and STLinkV2-1 firmwares both support this feature. When using a probe with target type auto-detection, you do not need to tell pyOCD

To check whether your debug probe supports auto-detection, run pyocd list. This command prints a list of all connected debug probes. If a probe does not support auto-detection, the name of the probe firmware is printed. Whereas probes that do support auto-detection will show up with the name of the board plus the target type in brackets.

Example probe listing:

  #   Probe                               Unique ID
  0   DISCO-L475VG-IOT01A [stm32l475xg]   066EFF555051897267233656
  1   KEIL - Tools By ARM Keil ULINKplus  L02932983A

This example shows two probes. The first does support auto-detection, while the second does not. Note how the first probe’s description shows the name of the board and the target type “stm32l475xg” in brackets.

If the target type is not auto-detected, it will default to “cortex_m” unless specified as described above.

Target support sources


PyOCD has built-in support for over 70 popular MCUs from various vendors.

Because new targets are addded fairly often, the best way to see the available built-in targets is by running pyocd list --targets. This command will print a table of supported target types. Built-in targets will be identified as such in the “SOURCE” column.


The CMSIS specification defines the Device Family Pack (DFP) standard for distributing device support files. PyOCD uses DFPs as a means to add support for devices that do not have support built in. This allows pyOCD to support nearly every Cortex-M MCU on the market. It also means that pyOCD can immediately support newly released devices as soon as the silicon vendor makes a DFP available.

The CMSIS-Pack list web page is the official list of all available packs. You can choose to download individual packs.

There are two methods to use a DFP within pyOCD. First, you can let pyOCD fully manage packs for you, where all you have to do is tell it the part number(s) for which you wish to install support. Alternatively, you can manually download packs and tell pyOCD which pack files to use.

Note: As with any software release, CMSIS-Packs may be buggy and have incorrect or missing information about a device. Sometimes memory regions will be missing, attributes of a memory region or flash algorithm might be wrong, or the flash algorithm itself might be buggy. If you encounter an issue with a pack, you can report the issue on the pyOCD GitHub site to let other users know about the problem. You should also report the issue to the pack vendor so it can be fixed.

Managed packs

The pyocd tool’s pack subcommand provides completely automated management of CMSIS-Packs. It allows you to install new device support with a single command line invocation.

As part of the pack management, pyOCD keeps an index of all available DFPs. The pack index is a two- level hierarchy of pack lists, with the top level pointing to individual vendor indexes. Once the index is downloaded, pyOCD can very quickly locate the DFP that provides support for a given MCU part number. To learn more about the index, see the CMSIS-Pack index files documentation.

The two most useful subcommands of the pack subcommand are find and install. The options accept a glob-style pattern that is matched against MCU part numbers in the index. If the index has not been downloaded yet, that will be done first, or you can run pyocd pack update yourself. The find subcommand will print out a list of matching part numbers and the names of the containing DFPs. install performs the same match, but downloads and installs the matching DFPs. The part number patterns are matched case-insensitively and as a contains comparison.

For instance, if you know the specific part number of the device you are using, say STM32L073, you can run this command to install support:

$ pyocd pack install stm32l073

This will download the index if required, then download the STM32L0xx_DFP pack. The

As another example, to find which pack(s) support the NXP MK26F family, you could run:

$ pyocd pack find k26

This will print a table similar to:

  Part             Vendor   Pack          Version
  MK26FN2M0CAC18   NXP      MK26F18_DFP   11.0.1
  MK26FN2M0VLQ18   NXP      MK26F18_DFP   11.0.1
  MK26FN2M0VMD18   NXP      MK26F18_DFP   11.0.1
  MK26FN2M0VMI18   NXP      MK26F18_DFP   11.0.1

Once a DFP is installed, the pyocd list --targets command will show the new targets in its output, and you can immediately begin using the target support with the other pyocd subcommands.

To get a list of all installed packs, use the pack show subcommand.

Manual pack usage

If you prefer to manually manage packs, or if the managed pack system cannot access online packs from your network, you can download them yourself from the official CMSIS-Pack list. The downloaded pack files can be stored wherever you like. Typically you would group them in a single common directory. Another good option is to place the pack file used with a project in a project-relative location.

There are two ways to use a manually downloaded pack.

The simplest option is to pass the --pack option to the pyocd tool, specifying the path to the .pack file. PyOCD does not cache any information about packs used this way, so this argument must be passed for every invocation in addition to the other arguments. For instance, to run the GDB server, you might execute pyocd gdbserver --pack=Keil.STM32L4xx_DFP.2.2.0.pack. Note that you can pass multiple --pack arguments to pyOCD, which might be useful in a scripted execution of pyOCD.

For a more permanent solution, use a pyocd.yaml configuration file. In the config file, set the pack session option to either a single .pack file path or a list of paths. Now when you run the pyocd tool, it will automatically pick up the pack file(s) to use.

Here is an example config file that lists two packs.

  - /Users/admin/CMSIS-Packs/Keil.STM32L0xx_DFP.2.0.0.pack
  - /Users/admin/CMSIS-Packs/NXP.MKV58F24_DFP.11.0.0.pack

To see the targets provided by a .pack file, run pyocd list --targets and pass the approprate --pack option or use a config file, as described above.

Note: .pack files are simply zip archives with a different extension. To examine the contents of a pack, change the extension to .zip and extract.