Small desktop app (Tauri) that runs a local OpenAI-compatible → Shroud proxy in-process (no Node.js on the user’s machine). Rust implementation: src-tauri/src/llm_proxy.rs. The 1claw proxy CLI mirrors this behavior in packages/cli/src/commands/proxy.ts in the 1claw monorepo.
- License: MIT (see
LICENSE) - Requirements: A 1Claw agent with Shroud enabled and
ocv_credentials (oruuid:ocv_…). Network access to your Vault API and Shroud URLs.
Source of truth: github.com/1clawAI/shroud-bridge.
git clone https://github.com/1clawAI/shroud-bridge.git
cd shroud-bridge
npm install
npm run devProduction bundle:
npm run buildInstallers appear under src-tauri/target/release/bundle/ (.app / .dmg on macOS, .msi / .exe on Windows, .deb / .AppImage on Linux depending on Tauri defaults).
The main repo vendors this package at packages/shroud-bridge:
git clone --recurse-submodules https://github.com/1clawAI/1claw.git
cd 1claw/packages/shroud-bridge
npm install
npm run devThe monorepo runs .github/workflows/shroud-bridge.yml on Ubuntu, Windows, and macOS when packages/shroud-bridge/ changes.
- Agent credentials are stored in browser localStorage inside the webview (optional persistence). Treat the machine as trusted.
- For production hardening, prefer OS keychain integration and/or short-lived tokens in a future version.
- Parity: When changing proxy behavior, update
src-tauri/src/llm_proxy.rsand the CLIproxycommand together (or add shared contract tests).