carbon-lang

Carbon

Purpose

Guide agents through the Carbon programming language: syntax fundamentals, bidirectional C++ interoperability via Carbon.h, building from source with carbon-toolchain, the experimental memory safety model (checked borrows), current pre-1.0 limitations, and when to evaluate Carbon versus staying on C++.

When to Use

• Evaluating Carbon for a greenfield project with heavy C++ dependencies
• Calling C++ libraries from Carbon or exposing Carbon to C++
• Experimenting with Carbon toolchain from GitHub source
• Understanding Carbon's migration path from C++ codebases
• Assessing readiness for production (currently experimental)
• Comparing Carbon ergonomics to modern C++ (C++20/23)

Workflow

1. Project status awareness

Carbon (2026 state)
├── Experimental — no 1.0 release
├── Focus: C++ interoperability and migration
├── No production stdlib equivalent to C++ yet
└── Toolchain under active development

Use Carbon for exploration and migration prototyping, not production systems without team acceptance of instability.

2. Build carbon-toolchain

git clone https://github.com/carbon-language/carbon-lang
cd carbon-lang

# Prerequisites: LLVM, Clang, CMake, Ninja
./scripts/run_bazelisk.py build \
  //toolchain:install \
  --symlink_prefix=carbon/

# Add to PATH
export PATH="$PWD/carbon/bin:$PATH"

carbon --version

# Online explorer (Compiler Explorer instance)
# http://carbon.compiler-explorer.com/

3. Basic syntax

package Sample api;

fn Add(a: i32, b: i32) -> i32 {
    return a + b;
}

fn Main() -> i32 {
    var x: i32 = 3;
    var y: i32 = 4;
    Print(ToString(Add(x, y)));
    return 0;
}

Carbon	C++ equivalent
fn	function
var x: i32	int32_t x
class / interface	class / pure virtual
impl	method definitions
package	namespace/module

4. C++ interop — calling C++ from Carbon

// math.h (C++ header)
#pragma once
namespace Math {
    int Add(int a, int b);
}