Files
Rohit PrasadandDevika Verma d53e058eda Introduce OpenWorker Harness SDK (in beta), agent service cluster and relevant developer documentation (#844)
Allow developers to easily build AI apps without worrying about the harness part.
Bundles the Rust agent runtime as a SDK (Rust and TypeScript) for building desktop and web applications. Applications can manage a local harness or connect to an authenticated harness service on k8s with separately scaled workers and sandbox execution.

The implementation is primarily contained in harness/, with deployment tooling in deploy/, Markdown guides in docs/harness/guides/, and the static documentation site in docs-site/. It does not replace the existing Python harness or migrate OpenWorker Desktop.

Co-authored-by: Devika Verma <devikaverma11@gmail.com>
2026-11-06 09:45:20 -08:00

1.4 KiB

Native tool integration example

A small SDK-capable executable demonstrating how an application links its own Rust tool crates into an OpenWorker host. It is not a desktop application.

cd harness
cargo build --release --locked -p native-tools-runtime
cargo test --locked -p notification-tools -p device-tools -p computer-tools --lib

Use target/release/native-tools-runtime (.exe on Windows) as the SDK-managed child executable. For embedded Rust, register the same crates in the registry passed to Harness::builder().tools(registry).

  • notification-tools: notifications.notify_user emits a UI notification event. A separately invoked macOS dialog example is opt-in and not part of normal tests.
  • device-tools: device.computer_status, device.audio_status, device.set_volume, device.open_settings use native Windows APIs; other platforms return unsupported, not simulated results.
  • computer-tools: bounded Windows screenshot/input operations with explicit access grants, observation identities and expiry. See that crate's source and computer-use boundary.

Registration does not expose every tool to every agent. Select qualified tool names in each agent/coworker definition. Native callbacks carry the host process's privileges; approval and OS sandboxing are separate concerns. Do not grant access to the desktop or change device settings in unattended test runs.