Use this SDK to add realtime video, audio and data features to your C++ app. By connecting to LiveKit Cloud or a self-hosted server, you can quickly build applications such as multi-modal AI, live streaming, or video calls with just a few lines of code.

• CMake ≥ 3.20
• Rust / Cargo (latest stable toolchain)
• Git LFS (required for examples) Some example data files (e.g., audio assets) are stored using Git LFS. You must install Git LFS before cloning or pulling the repo if you want to run the examples.
• livekit-cli install livekit-cli by following the (official livekit docs)[https://docs.livekit.io/intro/basics/cli/start/]
• livekit-server install livekit-server by following the (official livekit docs)[https://docs.livekit.io/transport/self-hosting/local/]

Platform-Specific Requirements:

• Windows: Visual Studio 2019+, vcpkg
• Linux: sudo apt install libprotobuf-dev libssl-dev (protobuf 3.x)
• macOS: brew install protobuf (protobuf 3.x)

• Windows: ✅ All dependencies included (DLLs bundled) - ready to use
• Linux: ⚠️ Requires libprotobuf and libssl-dev; deploy liblivekit_ffi.so with your executable
• macOS: ⚠️ Requires protobuf; deploy liblivekit_ffi.dylib with your executable

Note: If the SDK was built with Protobuf 6.0+, you also need libabsl-dev (Linux) or abseil (macOS).

Make sure to initialize the Rust submodule (client-sdk-rust):

# Option 1: Clone with submodules in one step
git clone --recurse-submodules https://github.com/livekit/client-sdk-cpp.git

# Option 2: Clone first, then initialize submodules
git clone https://github.com/livekit/client-sdk-cpp.git
cd client-sdk-cpp
git submodule update --init --recursive

Linux/macOS:

./build.sh clean          # Clean CMake build artifacts
./build.sh clean-all      # Deep clean (C++ + Rust + generated files)
./build.sh debug          # Build Debug version
./build.sh release        # Build Release version
./build.sh debug-tests    # Build Debug with tests
./build.sh release-tests  # Build Release with tests

Windows Using build scripts:

.\build.cmd clean          # Clean CMake build artifacts
.\build.cmd clean-all      # Deep clean (C++ + Rust + generated files)
.\build.cmd debug          # Build Debug version
.\build.cmd release        # Build Release version
.\build.cmd debug-tests    # Build Debug with tests
.\build.cmd release-tests  # Build Release with tests

cmake -S . -B build -DCMAKE_TOOLCHAIN_FILE="$PWD/vcpkg/scripts/buildsystems/vcpkg.cmake"  # Generate Makefiles in build folder
# Build (Release or Debug)
cmake --build build --config Release
# or:
cmake --build build --config Debug
# Clean CMake build artifacts
Remove-Item -Recurse -Force build

Note (Windows), This assumes vcpkg is checked out in the repo root at .\vcpkg. You must install protobuf via vcpkg (so CMake can find ProtobufConfig.cmake and protoc), for example:

.\vcpkg\vcpkg install protobuf:x64-windows

For more control and platform-specific builds, see the detailed instructions in README_BUILD.md.

Prerequisites (Windows only):

• Set VCPKG_ROOT environment variable pointing to your vcpkg installation

# Windows PowerShell
$env:VCPKG_ROOT = "C:\path\to\vcpkg"

Prerequisites (Linux/macOS):

• Install system dependencies (see above)

Quick start:

# Windows
cmake --preset windows-release
cmake --build --preset windows-release

# Linux
cmake --preset linux-release
cmake --build --preset linux-release

# macOS
cmake --preset macos-release
cmake --build --preset macos-release

📖 For complete build instructions, troubleshooting, and platform-specific notes, see README_BUILD.md

The Dockerfile COPYs folders/files required to build the CPP SDK into the image. NOTE: this has only been tested on Linux

docker build -t livekit-cpp-sdk . -f docker/Dockerfile
docker run -it --network host livekit-cpp-sdk:latest bash

NOTE: if you are building your own Dockerfile, you will likely need to set the same ENV variables as in docker/Dockerfile, but to the relevant directories:

export CC=$HOME/gcc-14/bin/gcc
export CXX=$HOME/gcc-14/bin/g++
export LD_LIBRARY_PATH=$HOME/gcc-14/lib64:$LD_LIBRARY_PATH
export PATH=$HOME/.cargo/bin:$PATH
export PATH=$HOME/cmake-3.31/bin:$PATH

Before running any participant, create JWT tokens with the proper identity and room name, example

lk token create -r test -i your_own_identity  --join --valid-for 99999h --dev --room=your_own_room

./build/examples/SimpleRoom --url $URL --token <jwt-token>

You can also provide the URL and token via environment variables:

export LIVEKIT_URL=ws://localhost:7880
export LIVEKIT_TOKEN=<jwt-token>
./build/examples/SimpleRoom

End-to-End Encryption (E2EE) You can enable E2E encryption for the streams via --enable_e2ee and --e2ee_key flags, by running the following cmds in two terminals or computers. Note, jwt_token needs to be different identity

./build/examples/SimpleRoom --url $URL --token <jwt-token> --enable_e2ee --e2ee_key="your_key"

Note, all participants must use the exact same E2EE configuration and shared key. If the E2EE keys do not match between participants:

• Media cannot be decrypted
• Video tracks will appear as a black screen
• Audio will be silent
• No explicit error may be shown at the UI level

Press Ctrl-C to exit the example.

The SimpleRpc example demonstrates how to:

• Connect multiple participants to the same LiveKit room
• Register RPC handlers (e.g., arrival, square-root, divide, long-calculation)
• Send RPC requests from one participant to another
• Handle success, application errors, unsupported methods, and timeouts
• Observe round-trip times (RTT) for each RPC call

Before running any participant, create JWT tokens with caller, greeter and math-genius identities and room name.

lk token create -r test -i caller --join --valid-for 99999h --dev --room=your_own_room
lk token create -r test -i greeter --join --valid-for 99999h --dev --room=your_own_room
lk token create -r test -i math-genius --join --valid-for 99999h --dev --room=your_own_room

Every participant is run as a separate terminal process, note --role needs to match the token identity.

./build/examples/SimpleRpc --url $URL --token <jwt-token> --role=math-genius

The caller will automatically:

• Wait for the greeter and math-genius to join
• Perform RPC calls
• Print round-trip times
• Annotate expected successes or expected failures

• The SimpleDataStream example demonstrates how to:
• Connect multiple participants to the same LiveKit room
• Register text stream and byte stream handlers by topic (e.g. "chat", "files")
• Send a text stream (chat message) from one participant to another
• Send a byte stream (file/image) from one participant to another
• Attach custom stream metadata (e.g. sent_ms) via stream attributes
• Measure and print one-way latency on the receiver using sender timestamps
• Receive streamed chunks and reconstruct the full payload on the receiver

Before running any participant, create JWT tokens with caller and greeter identities and your room name.

lk token create -r test -i caller  --join --valid-for 99999h --dev --room=your_own_room
lk token create -r test -i greeter --join --valid-for 99999h --dev --room=your_own_room

Start the receiver first (so it registers stream handlers before messages arrive):

./build/examples/SimpleDataStream --url $URL --token <jwt-token>

On another terminal or computer, start the sender

./build/examples/SimpleDataStream --url $URL --token <jwt-token>

Sender (e.g. greeter)

• Waits for the peer, then sends a text stream ("chat") and a file stream ("files") with timestamps and metadata, logging stream IDs and send times.

Receiver (e.g. caller)

• Registers handlers for text and file streams, logs stream events, computes one-way latency, and saves the received file locally.

The SDK uses spdlog internally but does not expose it in public headers. All log output goes through a thin public API in <livekit/logging.h>.

Set once when you configure CMake. Calls below this threshold are completely removed by the preprocessor -- no format-string evaluation, no function call.

# Development (default): keep everything available
cmake -DLIVEKIT_LOG_LEVEL=TRACE ..

# Release: strip TRACE / DEBUG / INFO
cmake -DLIVEKIT_LOG_LEVEL=WARN ..

# Production: only ERROR and CRITICAL survive
cmake -DLIVEKIT_LOG_LEVEL=ERROR ..

Valid values: TRACE, DEBUG, INFO, WARN, ERROR, CRITICAL, OFF.

Among the levels that survived compilation you can still filter at runtime without rebuilding:

#include <livekit/livekit.h>

livekit::initialize();                           // default level: Info
livekit::setLogLevel(livekit::LogLevel::Debug);  // show more detail
livekit::setLogLevel(livekit::LogLevel::Warn);   // suppress info chatter

Replace the default stderr sink with your own handler. This is the integration point for frameworks like ROS2 (RCLCPP_* macros), Android logcat, or any structured-logging pipeline:

#include <livekit/livekit.h>

livekit::initialize();
livekit::setLogLevel(livekit::LogLevel::Trace);

livekit::setLogCallback(
    [](livekit::LogLevel level,
       const std::string &logger_name,
       const std::string &message) {
      // Route to your framework, e.g.:
      //   RCLCPP_INFO(get_logger(), "[%s] %s", logger_name.c_str(), message.c_str());
      myLogger.log(level, logger_name, message);
    });

// Pass nullptr to restore the default stderr sink:
livekit::setLogCallback(nullptr);

See examples/logging_levels/custom_sinks.cpp for three copy-paste-ready patterns: file logger, JSON structured lines, and a ROS2 bridge that maps LogLevel to RCLCPP_* macros.

The SDK includes integration and stress tests using Google Test (gtest).

Linux/macOS:

./build.sh debug-tests      # Build Debug with tests
./build.sh release-tests    # Build Release with tests

Windows:

.\build.cmd debug-tests
.\build.cmd release-tests

After building, run tests using ctest or directly:

# Run all tests via ctest
cd build-debug
ctest --output-on-failure

# Or run test executables directly
./build-debug/bin/livekit_integration_tests
./build-debug/bin/livekit_stress_tests

# Run specific test suites
./build-debug/bin/livekit_integration_tests --gtest_filter="*Rpc*"
./build-debug/bin/livekit_stress_tests --gtest_filter="*MaxPayloadStress*"

RPC integration and stress tests require a LiveKit server and two participant tokens:

# Required
export LIVEKIT_URL="wss://your-server.livekit.cloud"
export LIVEKIT_CALLER_TOKEN="<token with caller identity>"
export LIVEKIT_RECEIVER_TOKEN="<token with receiver identity>"

# Optional (for stress tests)
export RPC_STRESS_DURATION_SECONDS=3600   # Test duration (default: 1 hour)
export RPC_STRESS_CALLER_THREADS=4        # Concurrent caller threads (default: 4)

Generate tokens for RPC tests:

lk token create -r test -i rpc-caller --join --valid-for 99999h --dev --room=rpc-test-room
lk token create -r test -i rpc-receiver --join --valid-for 99999h --dev --room=rpc-test-room

• SDK Initialization: Initialize/shutdown lifecycle
• Room: Room creation, options, connection
• Audio Frame: Frame creation, manipulation, edge cases
• RPC: Round-trip calls, max payload (15KB), timeouts, errors, concurrent calls
• Stress Tests: High throughput, bidirectional RPC, memory pressure

brew install cmake protobuf rust

sudo apt update
sudo apt install -y cmake protobuf-compiler build-essential
curl https://sh.rustup.rs -sSf | sh

cd client-sdk-cpp
git fetch origin
git switch -c try-rust-main origin/main

# Sync submodule URLs and check out what origin/main pins (recursively):
git submodule sync --recursive
git submodule update --init --recursive --checkout

# Now, in case the nested submodule under yuv-sys didn’t materialize, force it explicitly:
cd ..
git -C client-sdk-rust/yuv-sys submodule sync --recursive
git -C client-sdk-rust/yuv-sys submodule update --init --recursive --checkout

# Sanity check:
git submodule status --recursive

cargo clean -p yuv-sys
cargo build -p yuv-sys -vv

In some cases, you may need to perform a full clean that deletes all build artifacts from both the Rust and C++ folders:

./build.sh clean-all

CPP SDK is using clang C++ format

brew install clang-format