mbed CLI


mbed CLI is the name of the ARM mbed command-line tool, packaged as mbed-cli. mbed CLI enables Git- and Mercurial-based version control, dependencies management, code publishing, support for remotely hosted repositories (GitHub, GitLab and mbed.org), use of the ARM mbed OS build system and export functions and other operations.

This document covers the installation and usage of mbed CLI.

Table of Contents

  1. Using mbed CLI
  2. Installing and uninstalling
  3. Understanding working context and program root
  4. Creating and importing programs
    1. Creating a new program
    2. Importing an existing program
  5. Adding and removing libraries
  6. Compiling code
    1. Toolchain selection
    2. Compiling your program
    3. Compiling static libraries
    4. Compile configuration system
    5. Compile-time customizations
  7. Exporting to desktop IDEs
  8. Testing
    1. Finding available tests
    2. Compiling and running tests
    3. Limiting the test scope
    4. Test directory structure
  9. Publishing your changes
    1. Checking status
    2. Pushing upstream
  10. Updating programs and libraries
    1. Updating to an upstream version
    2. Update examples
  11. mbed CLI configuration
  12. Troubleshooting

Using mbed CLI

The basic workflow for mbed CLI is to:

  1. Initialize a new repository, for either a new application (or library) or an imported one. In both cases, this action also adds the mbed OS codebase.
  2. Build the application code.
  3. Test your build.
  4. Publish your application.

To support long-term development, mbed CLI offers source control, including selective updates of libraries and the codebase, support for multiple toolchains and manual configuration of the system.

Tip: To list all mbed CLI commands, use mbed --help. A detailed command-specific help is available by using mbed <command> --help.


Windows, Linux and Mac OS X support mbed CLI. We’re keen to learn about your experience with mbed CLI on other operating systems at the mbed CLI development page.


  • Python - mbed CLI is a Python script, so you’ll need Python to use it. We test mbed CLI with version 2.7.11 of Python. It is not compatible with Python 3.

  • Git and Mercurial - mbed CLI supports both Git and Mercurial repositories, so you need to install both:

    • Git - version 1.9.5 or later.
    • Mercurial - version 2.2.2 or later.

The directories of Git and Mercurial executables (git and hg) need to be in your system’s PATH.

  • Command-line compiler or IDE toolchain - mbed CLI invokes the mbed OS 5 tools for various features, such as compiling, testing and exporting to industry standard toolchains. To compile your code, you need either a compiler or an IDE:
    • Compilers: GCC ARM, ARM Compiler 5, IAR.
    • IDE: Keil uVision, DS-5, IAR Workbench.

Video tutorial for manual installation

Video tutorial

Installing mbed CLI

You can get the latest stable version of mbed CLI through pip by running:

$ pip install mbed-cli

On Linux or Mac, you may need to run with sudo.

Alternatively, you can get the development version of mbed CLI by cloning the development repository https://github.com/ARMmbed/mbed-cli:

$ git clone https://github.com/ARMmbed/mbed-cli

Once cloned, you can install mbed CLI as a python package:

$ python setup.py install

On Linux or Mac, you may need to run with sudo.

Note: mbed CLI is compatible with Virtual Python Environment (virtualenv). You can read more about isolated Python virtual environments here.

Uninstalling mbed CLI

To uninstall mbed CLI, run:

pip uninstall mbed-cli

Adding Bash tab completion

To install mbed-cli bash tab completion navigate to the tools/bash_completion directory. Then copy the mbed script into your /etc/bash_completion.d/ or /usr/local/etc/bash_completion.d directory and reload your terminal.

Full documentation here

Quickstart video

Video tutorial

Before you begin: understanding the working context and program root

mbed CLI uses the current directory as a working context, in a similar way to Git, Mercurial and many other command-line tools. This means that before calling any mbed CLI command, you must first change to the directory containing the code you want to act on. For example, if you want to update the mbed OS sources in your mbed-example-program directory:

$ cd mbed-example-program
$ cd mbed-os
$ mbed update master   # This will update "mbed-os", not "my-program"

Various mbed CLI features require a program root, which should be under version control - either Git or Mercurial. This makes it possible to switch between revisions of the whole program and its libraries, control the program history, synchronize the program with remote repositories, share it with others and so on. Version control is also the primary and preferred delivery mechanism for mbed OS source code, which allows everyone to contribute to mbed OS.

Warning: mbed CLI stores information about libraries and dependencies in reference files that use the .lib extension (such as lib_name.lib). Although these files are human-readable, we strongly advise that you don’t edit these manually - let mbed CLI manage them instead.

Creating and importing programs

mbed CLI can create and import programs based on both mbed OS 2 and mbed OS 5.

Creating a new program for mbed OS 5

When you create a new program, mbed CLI automatically imports the latest mbed OS release. Each release includes all the components: code, build tools and IDE exporters.

With this in mind, let’s create a new program (we’ll call it mbed-os-program):

$ mbed new mbed-os-program
[mbed] Creating new program "mbed-os-program" (git)
[mbed] Adding library "mbed-os" from "https://github.com/ARMmbed/mbed-os" at latest revision in the current branch
[mbed] Updating reference "mbed-os" -> "https://github.com/ARMmbed/mbed-os/#89962277c20729504d1d6c95250fbd36ea5f4a2d"

This creates a new folder “mbed-os-program”, initializes a new repository and imports the latest revision of the mbed-os dependency to your program tree.

Tip: You can instruct mbed CLI to use a specific source control management system or prevent source control management initialization, by using --scm [name|none] option.

Use mbed ls to list all the libraries imported to your program:

$ cd mbed-os-program
$ mbed ls -a
mbed-os-program (mbed-os-program)
`- mbed-os (https://github.com/ARMmbed/mbed-os#89962277c207)

Note: If you want to start from an existing folder in your workspace, you can use mbed new ., which initializes an mbed program, as well as a new Git or Mercurial repository in that folder.

Creating a new program for mbed OS 2

mbed CLI is also compatible with mbed OS 2 programs based on the mbed library, and it automatically imports the latest mbed library release if you use the --mbedlib option:

$ mbed new mbed-classic-program --mbedlib
[mbed] Creating new program "mbed-classic-program" (git)
[mbed] Adding library "mbed" from "https://mbed.org/users/mbed_official/code/mbed/builds" at latest revision in the current branch
[mbed] Downloading mbed library build "f9eeca106725" (might take a minute)
[mbed] Unpacking mbed library build "f9eeca106725" in "D:\Work\examples\mbed-classic-program\mbed"
[mbed] Updating reference "mbed" -> "https://mbed.org/users/mbed_official/code/mbed/builds/f9eeca106725"
[mbed] Couldn't find build tools in your program. Downloading the mbed 2.0 SDK tools...

Creating a new program without OS version selection

You can create plain (empty) programs, without either mbed OS 5 or mbed OS 2, by using the --create-only option.

Importing an existing program

Use mbed import to clone an existing program and all its dependencies to your machine:

$ mbed import https://github.com/ARMmbed/mbed-os-example-blinky
[mbed] Importing program "mbed-os-example-blinky" from "https://github.com/ARMmbed/mbed-os-example-blinky" at latest revision in the current branch
[mbed] Adding library "mbed-os" from "https://github.com/ARMmbed/mbed-os" at rev #dd36dc4228b5
$ cd mbed-os-example-blinky

mbed CLI also supports programs based on mbed OS 2, which it automatically detects and which do not require additional options:

$ mbed import https://mbed.org/teams/mbed/code/mbed_blinky/
[mbed] Importing program "mbed_blinky" from "https://mbed.org/teams/mbed/code/mbed_blinky" at latest revision in the current branch
[mbed] Adding library "mbed" from "http://mbed.org/users/mbed_official/code/mbed/builds" at rev #f9eeca106725
[mbed] Couldn't find build tools in your program. Downloading the mbed 2.0 SDK tools...
$ cd mbed-os-example-blinky

You can use the “import” command without specifying a full URL; mbed CLI adds a prefix (https://github.com/ARMmbed) to the URL if one is not present. For example, this command:

$ mbed import mbed-os-example-blinky

is equivalent to this command:

$ mbed import https://github.com/ARMmbed/mbed-os-example-blinky

Importing from a Git or GitHub clone

If you have manually cloned a Git repository into your workspace and you want to add all missing libraries, then you can use the deploy command:

$ mbed deploy
[mbed] Adding library "mbed-os" from "https://github.com/ARMmbed/mbed-os" at rev #dd36dc4228b5

Don’t forget to set the current directory as the root of your program:

$ mbed new .

Adding and removing libraries

While working on your code, you may need to add another library to your application or remove existing libraries.

Adding a new library to your program is not the same as cloning the repository. Don’t clone a library using hg or git; use mbed add to add the library. This ensures that all libraries and sublibraries are populated as well.

Removing a library from your program is not the same as deleting the library directory. mbed CLI updates and removes library reference files. Use mbed remove to remove the library; don’t remove its directory with rm.

Adding a library

Use mbed add to add the latest revision of a library:

$ mbed add https://developer.mbed.org/users/wim/code/TextLCD/

Use the URL#hash format to add a library from a URL at a specific revision hash:

$ mbed add https://developer.mbed.org/users/wim/code/TextLCD/#e5a0dcb43ecc

Specifying a destination directory

If you want to specify a directory to which to add your library, you can give an additional argument to add, which names that directory. For example, If you’d rather add the previous library in a directory called “text-lcd” (instead of TextLCD):

$ mbed add https://developer.mbed.org/users/wim/code/TextLCD/ text-lcd

Although mbed CLI supports this functionality, we don’t encourage it. Adding a library with a name that differs from its source repository can lead to confusion.

Removing a library

If at any point you decide that you don’t need a library any more, you can use mbed remove with the path of the library:

$ mbed remove text-lcd

Compiling code

Toolchain selection

After importing a program or creating a new one, you need to tell mbed CLI where to find the toolchains that you want to use for compiling your source tree.

There are multiple ways to configure toolchain locations:
mbed_settings.py file in the root of your program. The tools will automatically create this file if it doesn’t already exist.
The mbed CLI configuration.
Setting an environment variable.
Adding directory of the compiler binary to your PATH.

Methods for configuring toolchains that appear earlier in the above list override methods that appear later.

Through mbed_settings.py

Edit mbed_settings.py to set your toolchain:

  • To use the ARM Compiler toolchain, set ARM_PATH to the base directory of your ARM Compiler installation (example: C:\Program Files\ARM\armcc5.06). The recommended version of the ARM Compiler toolchain is 5.06.
  • To use the GCC ARM Embedded toolchain, set GCC_ARM_PATH to the binary directory of your GCC ARM installation (example: C:\Program Files\GNU Tools ARM Embedded\4.9 2015q2\bin). Use version 4.9 of GCC ARM Embedded; version 5.0 or any more recent version might be incompatible with the tools.
  • To use the IAR EWARM toolhain, set IAR_PATH to the base directory of your IAR installation. Use versions 7.80 of IAR EWARM; prior versions might be incompatible with the tools.

Because mbed_settings.py contains local settings (possibly relevant only to a single OS on a single machine), you should not check it into version control.

Through mbed CLI configuration

You can set the ARM Compiler 5 location via the command:

$ mbed config -G ARM_PATH "C:\Program Files\ARM"
[mbed] C:\Program Files\ARM now set as global ARM_PATH

The -G switch tells mbed CLI to set this as a global setting, rather than local for the current program.

Supported settings for toolchain paths are ARM_PATH, GCC_ARM_PATH and IAR_PATH.

You can see the active mbed CLI configuration via:

$ mbed config --list
[mbed] Global config:
ARM_PATH=C:\Program Files\ARM\armcc5.06
IAR_PATH=C:\Program Files\IAR Workbench 7.0\arm

[mbed] Local config (D:\temp\mbed-os-program):
No local configuration is set

More information about mbed CLI configuration is available in the configuration section of this document.

Setting environment variable

For each of the compilers, mbed compile checks a corresponding environment variable for the compiler’s location. The environment variables are as follows:
MBED_ARM_PATH: The path to the base directory of your ARM Compiler installation. This should be the directory containing the directory containing the binaries for armcc and friends.
MBED_IAR_PATH: The path to the base directory of your IAR EWARM Compiler installation. This should be one directory containing the directory containing the binaries for iccarm and friends.
MBED_GCC_ARM_PATH: The path to the binary* directory of your GCC ARM Embedded Compiler installation. This should be the directory containing the binaries for arm-none-eabi-gcc and friends.

Compiler detection through the PATH

If none of the above are configured, the mbed compile command will fall back to checking your PATH for an executable that is part of the compiler suite in question. This check is the same as a shell would perform to find the executable on the command-line. When mbed compile finds the executable it is looking for, it uses the location of that executable as the appropriate path except in the case of GCC, which will not use a path.

Compiling your program

Use the mbed compile command to compile your code:

$ mbed compile -t ARM -m K64F
Building project mbed-os-program (K64F, GCC_ARM)
Compile: aesni.c
Compile: blowfish.c
Compile: main.cpp
... [SNIP] ...
Compile: configuration_store.c
Link: mbed-os-program
Elf2Bin: mbed-os-program
| Module                     | .text | .data | .bss |
| Fill                       |   170 |     0 | 2294 |
| Misc                       | 36282 |  2220 | 2152 |
| core/hal                   | 15396 |    16 |  568 |
| core/rtos                  |  6751 |    24 | 2662 |
| features/FEATURE_IPV4      |    96 |     0 |   48 |
| frameworks/greentea-client |   912 |    28 |   44 |
| frameworks/utest           |  3079 |     0 |  732 |
| Subtotals                  | 62686 |  2288 | 8500 |
Allocated Heap: 65540 bytes
Allocated Stack: 32768 bytes
Total Static RAM memory (data + bss): 10788 bytes
Total RAM memory (data + bss + heap + stack): 109096 bytes
Total Flash memory (text + data + misc): 66014 bytes
Image: BUILD/K64F/GCC_ARM/mbed-os-program.bin

The arguments for compile are:

  • -m <MCU> to select a target. If detect or auto parameter is passed to -m, then mbed CLI detects the connected target.
  • -t <TOOLCHAIN> to select a toolchain (of those defined in mbed_settings.py, see above). The value can be ARM (ARM Compiler 5), GCC_ARM (GNU ARM Embedded) or IAR (IAR Embedded Workbench for ARM).
  • --source <SOURCE> to select the source directory. The default is . (the current directory). You can specify multiple source locations, even outside the program tree.
  • --build <BUILD> to select the build directory. Default: BUILD/ inside your program root.
  • --profile <PATH_TO_BUILD_PROFILE> to select a path to a build profile configuration file. Example: mbed-os/tools/profiles/debug.json.
  • --library to compile the code as a static .a/.ar library.
  • --config to inspect the runtime compile configuration (see below).
  • -S or --supported shows a matrix of the supported targets and toolchains.
  • -f or --flash to flash/program a connected target after successful compile.
  • -c to build from scratch, a clean build or rebuild.
  • -j <jobs> to control the compile processes on your machine. The default value is 0, which infers the number of processes from the number of cores on your machine. You can use -j 1 to trigger a sequential compile of source code.
  • -v or --verbose for verbose diagnostic output.
  • -vv or --very_verbose for very verbose diagnostic output.

You can find the compiled binary, ELF image, memory usage and link statistics in the BUILD subdirectory of your program.

For more information on build profiles, see our build profiles and toolchain profiles pages.

Compiling static libraries

You can build a static library of your code by adding the --library argument to mbed compile. Static libraries are useful when you want to build multiple applications from the same mbed OS codebase without having to recompile for every application. To achieve this:

  1. Build a static library for mbed-os.
  2. Compile multiple applications or tests against the static library:
$ mbed compile -t ARM -m K64F --library --no-archive --source=mbed-os --build=../mbed-os-build
Building library mbed-os (K64F, ARM)
Completed in: (47.4)s

$ mbed compile -t ARM -m K64F --source=mbed-os/TESTS/integration/basic --source=../mbed-os-build --build=../basic-out
Building project basic (K64F, ARM)
Compile: main.cpp
Link: basic
Elf2Bin: basic
Image: ../basic-out/basic.bin

$ mbed compile -t ARM -m K64F --source=mbed-os/TESTS/integration/threaded_blinky --source=../mbed-os-build --build=..\/hreaded_blinky-out
Building project threaded_blinky (K64F, ARM)
Compile: main.cpp
Link: threaded_blinky
Elf2Bin: threaded_blinky
Image: ../threaded_blinky-out/threaded_blinky.bin

Compile configuration system

The compile configuration system provides a flexible mechanism for configuring the mbed program, its libraries and the build target.

Inspecting the configuration

You can use mbed compile --config to view the configuration:

$ mbed compile --config -t GCC_ARM -m K64F

To display more verbose information about the configuration parameters, use -v:

$ mbed compile --config -t GCC_ARM -m K64F -v

It’s possible to filter the output of mbed compile --config by specifying one or more prefixes for the configuration parameters that mbed CLI displays. For example, to display only the configuration defined by the targets:

$ mbed compile --config -t GCC_ARM -m K64F --prefix target

You may use --prefix more than once. To display only the application and target configuration, use two --prefix options:

$ mbed compile --config -t GCC_ARM -m K64F --prefix target --prefix app

Compile-time customizations


You can specify macros in your command-line using the -D option. For example:

$ mbed compile -t GCC_ARM -m K64F -c -DUVISOR_PRESENT

Compiling in debug mode

To compile in debug mode (as opposed to the default develop mode), use --profile mbed-os/tools/profiles/debug.json in the compile command-line:

$ mbed compile -t GCC_ARM -m K64F --profile mbed-os/tools/profiles/debug.json

Tip: If you have files that you want to compile only in debug mode, put them in a directory called TARGET_DEBUG at any level of your tree (then use --profile as explained above).

Automating toolchain and target selection

Using mbed target <target> and mbed toolchain <toolchain>, you can set the default target and toolchain for your program. You won’t have to specify these every time you compile or generate IDE project files.

You can also use mbed target detect, which detects the connected target board and uses it as a parameter to every subsequent compile and export.

Exporting to desktop IDEs

If you need to debug your code, you can export your source tree to an IDE project file to use the IDE’s debugging facilities. mbed CLI supports exporting to Keil uVision, IAR Workbench, a Makefile using GCC ARM, Eclipse using GCC ARM and other IDEs.

For example, to export to uVision, run:

$ mbed export -i uvision -m K64F

mbed CLI creates a .uvprojx file in the projectfiles/uvision folder. You can open the project file with uVision.


Use the mbed test command to compile and run tests.

The arguments to test are:
-m <MCU> to select a target for the compilation. If detect or auto parameter is passed, then mbed CLI will attempt to detect the connected target and compile against it.
-t <TOOLCHAIN> to select a toolchain (of those defined in mbed_settings.py, see above), where toolchain can be either ARM (ARM Compiler 5), GCC_ARM (GNU ARM Embedded), or IAR (IAR Embedded Workbench for ARM).
--compile-list to list all the tests that can be built.
--run-list to list all the tests that can be run (they must be built first).
--compile to only compile the tests.
--run to only run the tests.
-n <TESTS_BY_NAME> to limit the tests built or run to a comma separated list (ex. test1,test2,test3).
--source <SOURCE> to select the source directory. Default is . (the current directory). You can specify multiple source locations, even outside the program tree.
--build <BUILD> to select the build directory. Default: BUILD/ inside your program.
--profile <PATH_TO_BUILD_PROFILE> to select a path to a build profile configuration file. Example: mbed-os/tools/profiles/debug.json.
-c or --clean to clean the build directory before compiling.
--test-spec <TEST_SPEC> to set the path for the test spec file used when building and running tests (the default path is the build directory).
-v or --verbose for verbose diagnostic output.
-vv or --very_verbose for very verbose diagnostic output.

Invoke mbed test:

$ mbed test -m K64F -t GCC_ARM
Building library mbed-build (K64F, GCC_ARM)
Building project GCC_ARM to TESTS-unit-myclass (K64F, GCC_ARM)
Compile: main.cpp
Link: TESTS-unit-myclass
Elf2Bin: TESTS-unit-myclass
| Module    | .text | .data | .bss |
| Fill      |   74  |   0   | 2092 |
| Misc      | 47039 |  204  | 4272 |
| Subtotals | 47113 |  204  | 6364 |
Allocated Heap: 65540 bytes
Allocated Stack: 32768 bytes
Total Static RAM memory (data + bss): 6568 bytes
Total RAM memory (data + bss + heap + stack): 104876 bytes
Total Flash memory (text + data + misc): 48357 bytes
Image: build\tests\K64F\GCC_ARM\TESTS\mbedmicro-rtos-mbed\mutex\TESTS-unit-myclass.bin
mbedgt: test suite report:
| target       | platform_name | test suite                      | result | elapsed_time (sec) | copy_method |
| K64F-GCC_ARM | K64F          | TESTS-unit-myclass              | OK     | 21.09              |    shell    |
mbedgt: test suite results: 1 OK
mbedgt: test case report:
| target       | platform_name | test suite         | test case           | passed | failed | result | elapsed_time (sec) |
| K64F-GCC_ARM | K64F          | TESTS-unit-myclass | TESTS-unit-myclass1 | 1      | 0      | OK     | 5.00               |
| K64F-GCC_ARM | K64F          | TESTS-unit-myclass | TESTS-unit-myclass2 | 1      | 0      | OK     | 5.00               |
| K64F-GCC_ARM | K64F          | TESTS-unit-myclass | TESTS-unit-myclass3 | 1      | 0      | OK     | 5.00               |
mbedgt: test case results: 3 OK
mbedgt: completed in 21.28 sec

You can find the compiled binaries and test artifacts in the BUILD/tests/<TARGET>/<TOOLCHAIN> directory of your program.

Finding available tests

You can find the tests that are available for building by using the --compile-list option:

$ mbed test --compile-list
Test Case:
    Name: TESTS-functional-test1
    Path: .\TESTS\functional\test1
Test Case:
    Name: TESTS-functional-test2
    Path: .\TESTS\functional\test2
Test Case:
    Name: TESTS-functional-test3
    Path: .\TESTS\functional\test3

You can find the tests that are available for running by using the --run-list option:

$ mbed test --run-list
mbedgt: test specification file '.\build\tests\K64F\ARM\test_spec.json' (specified with --test-spec option)
mbedgt: using '.\build\tests\K64F\ARM\test_spec.json' from current directory!
mbedgt: available tests for built 'K64F-ARM', location '.\build\tests\K64F\ARM'
        test 'TESTS-functional-test1'
        test 'TESTS-functional-test2'
        test 'TESTS-functional-test3'

Compiling and running tests

You can specify to only build the tests by using the --compile option:

$ mbed test -m K64F -t GCC_ARM --compile

You can specify to only run the tests by using the --run option:

$ mbed test -m K64F -t GCC_ARM --run

If you don’t specify any of these, mbed test will first compile all available tests and then run them.

Limiting the test scope

You can limit the scope of the tests built and run by using the -n option. This takes a comma-separated list of test names as an argument:

$ mbed test -m K64F -t GCC_ARM -n TESTS-functional-test1,TESTS-functional-test2

You can use the wildcard character * to run a group of tests that share a common prefix without specifying each test individually. For instance, if you only want to run the three tests TESTS-functional-test1, TESTS-functional-test2 and TESTS-functional-test3, but you have other tests in your project, you can run:

$ mbed test -m NUCLEO_F429ZI -t GCC_ARM -n TESTS-functional*

Note: Some shells expand the wildcard character * into file names that exist in your working directory. To prevent this behavior, please see your shell’s documentation.

Test directory structure

Test code must follow this directory structure:

 |- main.cpp            # Optional main.cpp with main() if it is an application module.
 |- pqr.lib             # Required libs
 |- xyz.lib
 |- mbed-os
 |  |- frameworks        # Test dependencies
 |  |  `_greentea-client # Greentea client required by tests.
 |  |...
 |  `- TESTS              # Tests directory. Special name upper case TESTS is excluded during application build process
 |     |- TestGroup1      # Test Group directory
 |     |  `- TestCase1    # Test case source directory
 |     |      `- main.cpp # Test source
 |     |- TestGroup2
 |     |   `- TestCase2
 |     |      `- main.cpp
 |     `- host_tests      # Python host tests script directory
 |        |- host_test1.py
 |        `- host_test2.py
 `- build                 # Build directory
     |- <TARGET>          # Target directory
     | `- <TOOLCHAIN>     # Toolchain directory
     |   |- TestCase1.bin # Test binary
     |   `- TestCase2.bin
     | ....

As shown above, tests exist inside TESTS\testgroup\testcase\ directories. Please note that TESTS is a special upper case directory that is excluded from module sources while compiling.

Note: mbed test does not work in applications that contain a main function that is outside of a TESTS directory.

Publishing your changes

Checking status

As you develop your program, you’ll edit parts of it. You can get the status of all the repositories in your program (recursively) by running mbed status. If a repository has uncommitted changes, this command displays these changes.

Here’s an example:

[mbed] Status for "mbed-os-program":
 M main.cpp
 M mbed-os.lib
?? gdb_log.txt
?? test_spec.json

[mbed] Status for "mbed-os":
 M tools/toolchains/arm.py
 M tools/toolchains/gcc.py

[mbed] Status for "mbed-client-classic":
 M source/m2mtimerpimpl.cpp

[mbed] Status for "mbed-mesh-api":
 M source/include/static_config.h

You can then commit or discard these changes through that repository’s version control system.

Pushing upstream

To push the changes in your local tree upstream, run mbed publish. mbed publish works recursively, pushing the leaf dependencies first, then updating the dependents and pushing them too.

Let’s assume that the list of dependencies of your program (obtained by running mbed ls) looks like this:

my-mbed-os-example (a5ac4bf2e468)
|- mbed-os (5fea6e69ec1a)
`- my-libs (e39199afa2da)
   |- my-libs/iot-client (571cfef17dd0)
   `- my-libs/test-framework (cd18b5a50df4)

Let’s assume that you make changes to iot-client. mbed publish detects the change on the leaf iot-client dependency and asks you to commit it. Then mbed publish detects that my-libs depends on iot-client, updates the my-libs dependency on iot-client to its latest version by updating the iot-client.lib file and asks you to commit it. This propagates up to my-libs and finally to your program, my-mbed-os-example.

Publishing a local program or library

When you create a new (local) version control managed program or library, its revision history exists only locally. The repository is not associated with the remote one. To publish the local repository, please follow these steps:

  1. Create a new empty repository on the remote site. This can be on a public repository hosting service (GitHub, Bitbucket, mbed.org), your own service or a different location on your system.
  2. Copy the URL/location of the new repository in your clipboard.
  3. Open command-line in the local repository directory (for example, change directory to mbed-os-example/local-lib).
  4. To associate the local repository:
    * For Git, run git remote add origin <url-or-paht-to-your-remote-repo>.
    * For Mercurial, edit .hg/hgrc and add (or replace if exists):

    [paths] default = <url-or-paht-to-your-remote-repo>

  5. Run mbed publish to publish your changes.

In a scenario with nested local repositories, start with the leaf repositories first.

Forking workflow

Git enables a workflow where the publish/push repository may be different than the original (“origin”) one. This allows new revisions in a fork repository while maintaining an association with the original repository. To use this workflow, first import an mbed OS program or mbed OS itself, and then associate the push remote with your fork. For example:

$ git remote set-url --push origin https://github.com/screamerbg/repo-fork

Both git commit & git push and mbed publish push the new revisions to your fork. You can fetch from the original repository using mbed update or git pull. If you explicitly want to fetch or pull from your fork, then you can use git pull https://github.com/screamerbg/repo-fork [branch].

Through the workflow explained above, mbed CLI maintains association to the original repository (which you may want to send a pull request to) and will record references with the revision hashes that you push to your fork. Until your pull request (PR) is accepted, all recorded references are invalid. Once the PR is accepted, all revision hashes from your fork become part the original repository, making them valid.

Updating programs and libraries

You can update programs and libraries on your local machine so that they pull in changes from the remote sources (Git or Mercurial).

As with any mbed CLI command, mbed update uses the current directory as a working context. Before calling mbed update, you should change your working directory to the one you want to update. For example, if you’re updating mbed-os, use cd mbed-os before you begin updating.

Tip: Synchronizing library references: Before triggering an update, you may want to synchronize any changes that you’ve made to the program structure by running mbed sync, which updates the necessary library references and removes the invalid ones.

Protection against overwriting local changes

The update command fails if there are changes in your program or library that mbed update could overwrite. This is by design. mbed CLI does not run operations that would result in overwriting uncommitted local changes. If you get an error, take care of your local changes (commit or use one of the options below), and then rerun mbed update.

Updating to an upstream version

Updating a program

To update your program to another upstream version, go to the root folder of the program, and run:

$ mbed update [branch|tag|revision]

This fetches new revisions from the remote repository, updating the program to the specified branch, tag or revision. If you don’t specificy any of these, then mbed update updates to the latest revision of the current branch. mbed update performs this series of actions recursively against all dependencies in the program tree.

Updating a library

You can change the working directory to a library folder and use mbed update to update that library and its dependencies to a different revision than the one referenced in the parent program or library. This allows you to experiment with different versions of libraries/dependencies in the program tree without having to change the parent program or library.

There are three additional options that modify how unpublished local libraries are handled:

  • mbed update --clean-deps - Update the current program or library and its dependencies, and discard all local unpublished repositories. Use this with caution because your local unpublished repositories cannot be restored unless you have a backup copy.

  • mbed update --clean-files - Update the current program or library and its dependencies, discard local uncommitted changes and remove any untracked or ignored files. Use this with caution because your local unpublished repositories cannot be restored unless you have a backup copy.

  • mbed update --ignore - Update the current program or library and its dependencies, and ignore any local unpublished libraries (they won’t be deleted or modified, just ignored).

Update examples

There are two main scenarios when updating:

  • Update with local uncommitted changes: dirty update.

Run mbed update [branch|revision|tag_name]. You might have to commit or stash your changes if the source control tool (Git or Mercurial) throws an error that the update will overwrite local changes.

  • Discard local uncommitted changes: clean update.

Run mbed update [branch|revision|tag_name] --clean

Specifying a branch to mbed update will only check out that branch and won’t automatically merge or fast-forward to the remote/upstream branch. You can run mbed update to merge (fast-forward) your local branch with the latest remote branch. On Git you can do git pull.

Warning: The --clean option tells mbed CLI to update that program or library and its dependencies and discard all local changes. This action cannot be undone; use with caution.

Combining update options

You can combine the options of the mbed update command for the following scenarios:

  • mbed update --clean --clean-deps --clean-files - Update the current program or library and its dependencies, remove all local unpublished libraries, discard local uncommitted changes and remove all untracked or ignored files. This wipes every single change that you made in the source tree and restores the stock layout.

  • mbed update --clean --ignore - Update the current program or library and its dependencies, but ignore any local repositories. mbed CLI will update whatever it can from the public repositories.

Use these with caution because your uncommitted changes and unpublished libraries cannot be restored.

mbed CLI configuration

You can streamline many options in mbed CLI with global and local configuration.

The mbed CLI configuration syntax is:

mbed config [--global] <var> [value] [--unset]
  • The global configuration (via --global option) defines the default behavior of mbed CLI across programs unless overridden by local settings.
  • The local configuration (without --global) is specific to mbed program and allows overriding of global or default mbed CLI settings.
  • If you do not specify a value, then mbed CLI prints the value for this setting in this context.
  • The --unset option allows you to remove a setting.
  • The --list option allows you to list global and local configuration.

Here is a list of configuration settings and their defaults:

  • target - defines the default target for compile, test and export; an alias of mbed target. Default: none.
  • toolchain - defines the default toolchain for compile and test; can be set through mbed toolchain. Default: none.
  • ARM_PATH, GCC_ARM_PATH, IAR_PATH - defines the path to ARM Compiler, GCC ARM and IAR Workbench toolchains. Default: none.
  • protocol - defines the default protocol used for importing or cloning of programs and libraries. The possible values are https, http and ssh. Use ssh if you have generated and registered SSH keys (Public Key Authentication) with a service such as GitHub, GitLab, Bitbucket and so on. Read more about SSH keys here. Default: https.
  • depth - defines the clone depth for importing or cloning and applies only to Git repositories. Note that though this option may improve cloning speed, it may also prevent you from correctly checking out a dependency tree when the reference revision hash is older than the clone depth. Read more about shallow clones here. Default: none.
  • cache - defines the local path that stores small copies of the imported or cloned repositories, and mbed CLI uses it to minimize traffic and speed up future imports of the same repositories. Use on or enabled to turn on caching in the system temp path. Use none to turn caching off. Default: none (disabled).


Unable to import Mercurial (mbed.org) programs or libraries.

  1. Check whether you have Mercurial installed in your system path by running hg in command prompt. If you’re receiving “command not found” or a similar message, then you need to install Mercurial, and add it to your system path.

  2. Try to clone a Mercurial repository directly. For example, hg clone https://developer.mbed.org/teams/mbed/code/mbed_blinky/. If you receive an error similar to abort: error: [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed (_ssl.:590), then your system certificates are out of date. You need to update your system certificates and possibly add the host certificate fingerprint of mbed.com and mbed.org. Read more about Mercurial’s certificate management here.

Various issues when running mbed CLI in Cygwin environment

Currently mbed CLI is not compatible with Cygwin environment and cannot be executed inside it (https://github.com/ARMmbed/mbed-cli/issues/299).