RayTrace is now on PyPI

raytrace-cli 0.11.1 is now available on PyPI. Install RayTrace with your Python tooling, then use the CLI to run coding agents in cloud or local sandboxes and review their traces, tool calls, cost, and execution evidence.

The Python package delivers the same CLI as @raytrace-cli/cli on npm and includes its own Node.js runtime. You do not need to install Node.js separately.

Install with your Python tools

Use Python 3.8 or newer. To install this release with pip, run:

python -m pip install --upgrade raytrace-cli==0.11.1
raytrace --version

If you manage command-line tools with uv or pipx, choose either:

uv tool install raytrace-cli==0.11.1

pipx install raytrace-cli==0.11.1

Start a sandbox

From your project directory, sign in and create a cloud sandbox:

raytrace auth login
raytrace sandbox create

Follow the Get started guide to connect and run your coding agent. For a sandbox on your own machine, use raytrace sandbox create --local. Local sandboxes also require Lima; the CLI offers to install it during creation.

Connect through MCP

You can also launch RayTrace's MCP server through uv without a separate global CLI installation. For Claude Code:

claude mcp add raytrace -- uvx --from raytrace-cli raytrace mcp

Other MCP clients can use the same uvx --from raytrace-cli raytrace mcp command. See the CLI documentation for client configuration.

Supported platforms

Prebuilt packages are available for:

• macOS 13.5+: Apple Silicon and Intel.
• Linux with glibc 2.28+: x86-64 and ARM64.
• Windows: x64 and ARM64.

See the 0.11.1 release on PyPI for package details and downloads.