The Crazyflie Python API explanation

In order to easily use and control the Crazyflie there's a library made in Python that gives high-level functions and hides the details. This page contains generic information about how to use this library and the API that it implements.

If you are interested in more details look in the PyDoc in the code or:

Structure of the library

The library is asynchronous and based on callbacks for events. Functions like open_link will return immediately, and the callback connected will be called when the link is opened. The library doesn't contain any threads or locks that will keep the application running, it's up to the application that is using the library to do this.

There are a few synchronous wrappers for selected classes that create a synchronous API by wrapping the asynchronous classes, see the Synchronous API section

Uniform Resource Identifier (URI)

All communication links are identified using an URI built up of the following: InterfaceType://InterfaceId/InterfaceChannel/InterfaceSpeed

Currently we have radio, serial, usb, debug, udp interfaces. Here are some examples:

  • radio://0/10/2M : Radio interface, USB dongle number 0, radio channel 10 and radio speed 2 Mbit/s: radio://0/10/2M
  • debug://0/1 : Debug interface, id 0, channel 1
  • usb://0 : USB cable to microusb port, id 0
  • serial://ttyAMA0 : Serial port, id ttyAMA0
  • tcp://aideck-AABBCCDD.local:5000 : TCP network connection, Name: aideck-AABBCCDD.local, port 5000

Variables and logging

The library supports setting up logging configurations that are used for logging variables from the firmware. Each log configuration contains a number of variables that should be logged as well as a time period (in ms) of how often the data should be sent back to the host. Once the log configuration is added to the firmware the firmware will automatically send back the data at every period. These configurations are used in the following way:

  • Connect to the Crazyflie (log configurations needs a TOC to work)
  • Create a log configuration that contains a number of variables to log and a period at which they should be logged
  • Add the log configuration. This will also validate that the log configuration is sane (i.e uses a supported period and all variables are in the TOC)
  • After checking that the configuration is valid, set up callbacks for the data in your application and start the log configuration
  • Each time the firmware sends data back to the host, the callback will be called with a time-stamp and the data

There's are few limitations that need to be taken into account:

  • The maximum length for a log packet is 26 bytes. This, for for example, allows to log 6 floats and one uint16_t (6*4 + 2 bytes) in a single packet.
  • The minimum period of a for a log configuration is multiples of 10ms


The library supports reading and writing parameters at run-time to the firmware. This is intended to be used for data that is not continuously changed by the firmware, like setting regulation parameters and reading out if the power-on self-tests passed. Parameters should only be changed in the firmware when being set from the host (cfclient or a cflib script) or during start-up.

The library doesn't continuously update the parameter values, this should only be done once after connecting. After each write to a parameter the firmware will send back the updated value and this will be forwarded to callbacks registered for reading this parameter.

The parameters should be used in the following way:

  • Register parameter updated callbacks at any time in your application
  • Connect to your Crazyflie (this will download the parameter TOC)
  • Request updates for all the parameters
  • The library will call all the callbacks registered
  • The host can now write parameters that will be forwarded to the firmware
  • For each write all the callbacks registered for this parameter will be called back

There is an exception for experimental support to change the parameter from within firmware’s app layer. However do mind that this functionality is not according to the design of the parameters framework so that the host might not be updated correctly on the parameter change.

Variable and parameter names

All names of parameters and log variables use the same structure:

The group should be used to bundle together logical groups, like everything that deals with the stabilizer should be in the group stabilizer.

There's a limit of 28 chars in total and here are some examples:

  • stabilizer.roll
  • stabilizer.pitch
  • pm.vbat
  • imu_tests.MPU6050
  • pid_attitide.pitch_kd



All callbacks are handled using the Caller class that contains the following methods:

        """ Register cb as a new callback. Will not register duplicates. """

        """ Un-register cb from the callbacks """

        """ Call the callbacks registered with the arguments args """

Before the library can be used the link drivers have to he initialized. This will search for available drivers and instantiate them.

       """ Search for and initialize link drivers."""

Serial driver

The serial driver is disabled by default and has to be enabled to be used. Enable it in the call to init_drivers().


Operations on the link and connection will return immediately and will call the following callbacks when events occur:

    # Called on disconnect, no matter the reason
    disconnected = Caller()
    # Called on unintentional disconnect only
    connection_lost = Caller()
    # Called when the first packet in a new link is received
    link_established = Caller()
    # Called when the user requests a connection
    connection_requested = Caller()
    # Called when the link is established and the TOCs (that are not cached)
    # have been downloaded
    connected = Caller()
    # Called when the the link is established and all data, including parameters have been downloaded
    fully_connected = Caller()
    # Called if establishing of the link fails (i.e times out)
    connection_failed = Caller()
    # Called for every packet received
    packet_received = Caller()
    # Called for every packet sent
    packet_sent = Caller()
    # Called when the link driver updates the link quality measurement
    link_quality_updated = Caller()

To register for callbacks the following is used:

    crazyflie = Crazyflie()

Finding a Crazyflie and connecting

The first thing to do is to find a Crazyflie quadcopter that we can connect to. This is done by queuing the library that will scan all the available interfaces (currently the debug and radio interface).

    available = cflib.crtp.scan_interfaces()
    for i in available:
        print "Interface with URI [%s] found and name/comment [%s]" % (i[0], i[1])

Opening and closing a communication link is done by using the Crazyflie object:

    crazyflie = Crazyflie()

Then you can use the following to close the link again:


Sending control setpoints with the commander framework

The control setpoints are not implemented as parameters, instead they have a special API.

Attitude Setpoints

    def send_setpoint(self, roll, pitch, yawrate, thrust):

To send a new control set-point use the following:

    roll    = 0.0
    pitch   = 0.0
    yawrate = 0
    thrust  = 10001
    crazyflie.commander.send_setpoint(roll, pitch, yawrate, thrust)

Thrust is an integer value ranging from 10001 (next to no power) to 60000 (full power). It corresponds to the mean thrust that will be applied to the motors. There is a battery compensation algorithm applied to make the thrust mostly independent of battery voltage. Roll/pitch are in degree and yawrate is in degree/seconds.

This command will set the attitude controller setpoint for the next 500ms. After 500ms without next setpoint, the Crazyflie will apply a setpoint with the same thrust but with roll/pitch/yawrate = 0, this will make the Crazyflie stop accelerating. After 2 seconds without new setpoint the Crazyflie will cut power from the motors.

Note that this command implements a motor lock mechanism that is intended to avoid flyaway when connecting a gamepad. You must send one command with thrust = 0 in order to unlock the command. This unlock procedure needs to be repeated if the watchdog described above kicks-in.

This the current default behavior, but it depends on the Supervisor configuration in the Crazyflie firmware.

Note: If you’d like to send rate commands instead, make sure that the flightmode.stabModeRoll/Pitch/Yaw is set to rate (0).

Other commander setpoints sending

If your Crazyflie has a positioning system (Loco, flowdeck, MoCap, Lighthouse), you can also send velocity or position setpoints, like for instance:

send_hover_setpoint(self, vx, vy, yawrate, zdistance)

Check out the automated API documentation for the Crazyflie cflib’s commander framework to find out what other functions you can use.


The parameter framework is used to read and set parameters. This functionality should be used when:

  • The parameter is not changed by the Crazyflie but by the client
  • The parameter is not read periodically

If this is not the case then the logging framework should be used instead.

To set a parameter you must be connected to the Crazyflie. A parameter is set using:

    param_name = ""
    param_value = 3
    crazyflie.param.set_value(param_name, param_value)

The parameter reading is done using callbacks. When a parameter is updated from the host (using the code above) the parameter will be read back by the library and this will trigger the callbacks. Parameter callbacks can be added at any time (you don't have to be connected to a Crazyflie).

    add_update_callback(group, name=None, cb=None)
        Add a callback for a specific parameter name or group. If name is not specified then
        all parameters in the group will trigger the callback. This callback will be executed
        when a new value is read from the Crazyflie.

        """ Request an update of the value for the supplied parameter. """

    set_value(complete_name, value)
        """ Set the value for the supplied parameter. """

Here's an example of how to use the calls.

    crazyflie.param.add_update_callback(group="group", name="name", param_updated_callback)

    def param_updated_callback(name, value):
        print "%s has value %d" % (name, value)

It is also possible to get the current value of a parameter (when connected) without using a callback

    value = get_value(complete_name)

Note 1 If you call set_value() and then directly call get_value() for a parameter, you might not read back the new value, but get the old one instead. The process is asynchronous and get_value() will not return the new value until the parameter value has propagated to the Crazyflie and back. Use the callback method if you need to be certain that you get the correct value after an update.

Note 2: get_value() and set_value() can not be called from callbacks until the Crazyflie is fully connected. Most notably they can not be called from the connected callback as the parameter values have not been downloaded yet. Use the fully_connected callback to make sure the system is ready for parameter use. It is OK to call get_value() and set_value() from the fully_connected callback.


The logging framework is used to enable the "automatic" sending of variable values at specified intervals to the client. This functionality should be used when:

  • The variable is changed by the Crazyflie and not by the client
  • The variable is updated at high rate and you want to read the value periodically

If this is not the case then the parameter framework should be used instead.

The API to create and get information from LogConfig:

    # Called when new logging data arrives
    data_received_cb = Caller()
    # Called when there's an error
    error_cb = Caller()
    # Called when the log configuration is confirmed to be started
    started_cb = Caller()
    # Called when the log configuration is confirmed to be added
    added_cb = Caller()

    add_variable(name, fetch_as=None)
        """Add a new variable to the configuration.

        name - Full name of the variable in the form
        fetch_as - String representation of the type the variable should be
                   fetched as (i.e uint8_t, float, FP16, etc)

        If no fetch_as type is supplied, then the stored type will be used
        (i.e the type of the fetched variable is the same as it's stored in the

        """Start the logging for this entry"""

        """Stop the logging for this entry"""

        """Delete this entry in the Crazyflie"""

The API for the log in the Crazyflie:

        """Add a log configuration to the logging framework.

        When doing this the contents of the log configuration will be validated
        and listeners for new log configurations will be notified. When
        validating the configuration the variables are checked against the TOC
        to see that they actually exist. If they don't then the configuration
        cannot be used. Since a valid TOC is required, a Crazyflie has to be
        connected when calling this method, otherwise it will fail."""

To create a logging configuration the following can be used:

    logconf = LogConfig(name="Logging", period_in_ms=100)
    logconf.add_variable("group1.name1", "float")
    logconf.add_variable("group1.name2", "uint8_t")
    logconf.add_variable("group2.name1", "int16_t")

The datatype is the transferred datatype, it will be converted from internal type to transferred type before transfers:

  • float
  • uint8_t and int8_t
  • uint16_t and int16_t
  • uint32_t and int32_t
  • FP16: 16bit version of floating point, allows to pack more variables in one packet at the expense of precision.

The logging cannot be started until your are connected to a Crazyflie:

    # Callback called when the connection is established to the Crazyflie
    def connected(link_uri):

        if logconf.valid:
            print "One or more of the variables in the configuration was not found in log TOC. No logging will be possible."

    def data_received_callback(timestamp, data, logconf):
        print "[%d][%s]: %s" % (timestamp,, data)

    def logging_error(logconf, msg):
        print "Error when logging %s" %

The values of log variables are transferred from the Crazyflie using CRTP packets, where all variables belonging to one logging configuration are transferred in the same packet. A CRTP packet has a maximum data size of 30 bytes, which sets an upper limit to the number of variables that can be used in one logging configuration. If the desired log variables do not fit in one logging configuration, a second configuration may be added.

    crazyflie.log.add_config([logconf1, logconfig2])

Synchronous API

The synchronous classes are wrappers around the asynchronous API, where the asynchronous calls/callbacks are replaced with blocking calls. The synchronous API does not provide the full flexibility of the asynchronous API, but is useful when writing small scripts, for logging for instance.

The synchronous API uses the python Context manager concept, that is the with keyword. A resource is allocated when entering a with section and automatically released when exiting it, for instance a connection or take off/landing of a Crazyflie.


The SyncCrazyflie class wraps a Crazyflie instance and mainly simplifies connect/disconnect.

Basic usage

    with SyncCrazyflie(uri) as scf:
        # A Crazyflie instance is created and is now connected. If the connection fails,
        # an exception is raised.

        # The underlying crazyflie object can be accessed through the cf member'kalman.resetEstimation', '1')

        # Do useful stuff
        # When leaving the "with" section, the connection is automatically closed

If some special properties are required for the underlying Crazyflie object, a Crazyflie instance can be passed in to the SyncCrazyflie instance.

    my_cf = Crazyflie(rw_cache='./cache')
    with SyncCrazyflie(uri, cf=my_cf) as scf:
        # The my_cf is now connected
        # Do useful stuff
        # When leaving the "with" section, the connection is automatically closed


The SyncLogger class wraps setting up, as well as starting/stopping logging. It works both for Crazyflie and SyncCrazyflie instances. To get the log values, iterate the instance.

    # Connect to a Crazyflie
    with SyncCrazyflie(uri) as scf:
        # Create a log configuration
        log_conf = LogConfig(name='myConf', period_in_ms=200)
        log_conf.add_variable('stateEstimateZ.vx', 'int16_t')

        # Start logging
        with SyncLogger(scf, log_conf) as logger:
            # Iterate the logger to get the values
            count = 0
            for log_entry in logger:
                # Do useful stuff
                count += 1
                if (count > 10):
                    # The logging will continue until you exit the loop
            # When leaving this "with" section, the logging is automatically stopped
        # When leaving this "with" section, the connection is automatically closed


The MotionCommander class is intended to simplify basic autonomous flight, where the motion control is done from the host computer. The Crazyflie takes off and makes when entering the “with” section, and lands when exiting. It has functions for basic movements that are blocking until the motion is finished.

The MotionCommander is using velocity set points and does not have a global coordinate system, all positions are relative. It is mainly intended to be used with a Flow deck.

    with SyncCrazyflie(URI) as scf:
        # We take off when the commander is created
        with MotionCommander(scf) as mc:
            # Move one meter forward
            # Move one meter back
            # The Crazyflie lands when leaving this "with" section
        # When leaving this "with" section, the connection is automatically closed


The PositionHlCommander is intended to simplify basic autonomous flight, where all the high level commands exist inside the Crazyflie firmware. The Crazyflie takes off when entering the “with” section, and lands when exiting. It has functions for basic movements that are blocking until the motion is finished.

The PositionHlCommander uses the high level commander in the Crazyflie and is based on a global coordinate system and absolute positions. It is intended to be used with a positioning system such as Loco, the lighthouse or a mocap system.

    with SyncCrazyflie(URI) as scf:
        with PositionHlCommander(scf, controller=PositionHlCommander.CONTROLLER_PID) as pc:
            # Go to the coordinate (0, 0, 1)
            pc.go_to(0.0, 0.0, 1.0)
            # The Crazyflie lands when leaving this "with" section
        # When leaving this "with" section, the connection is automatically closed