Summary

ClawLess addresses the need for serverless, browser-based execution of AI agents. It provides a secure, sandboxed Node.js runtime environment using WebContainers (WASM), enabling users to run, observe, and control AI agents entirely within their browser without requiring a backend server. This offers a significant benefit for developers and researchers seeking isolated, auditable, and controllable AI agent development and deployment.

How It Works

The core of ClawLess is its utilization of the WebContainer API, which runs a full Node.js environment compiled to WebAssembly (WASM) directly in the browser. This allows seamless access to the vast npm ecosystem (3.4 million+ packages). The system enforces strict security through a fully isolated WASM sandbox, preventing access to the host system. Key architectural components include a declarative YAML policy engine for granular control over agent actions (file access, network, processes), and comprehensive end-to-end audit logging that captures every interaction for compliance and debugging.

Quick Start & Requirements

• Primary install/run command:
  • Local development:

    git clone https://github.com/open-gitagent/clawless.git
    cd clawless
    npm install
    npm run dev
    

  • As a dependency:

    npm install clawcontainer
    

• Non-default prerequisites: API keys for supported AI providers (Anthropic, OpenAI, Google) are required for runtime functionality.
• Links: No direct links to official quick-start guides, docs, or demos were provided in the README text, beyond the repository itself.

Highlighted Details

• WebContainer-powered sandboxed runtime (WASM) providing OS-level isolation within the browser.
• Integrated Monaco Editor with multi-file tab support and an xterm.js terminal offering full PTY emulation.
• Declarative YAML policy engine for defining and enforcing agent guardrails, including file access, process execution, and network restrictions.
• Comprehensive end-to-end audit logging for all container activities: process lifecycle, file I/O, network requests, and Git events.
• GitHub integration enables cloning repositories and pushing changes directly from the browser environment.
• Supports multiple AI providers (Anthropic, OpenAI, Google) with configurable model selection.
• Extensible via a plugin system with lifecycle hooks and a template system for agent bootstrapping.

Maintenance & Community

The project is noted as being "Built with care by Shreyas Kapale / Lyzr." Community interaction points mentioned include GitHub Discussions and Issues, though direct links were not provided in the text. Users are encouraged to star the repository.

Licensing & Compatibility

The license type is not explicitly stated in the provided README text. Compatibility is primarily for browser-based execution using standard web technologies.

Limitations & Caveats

Runtime state is persisted to localStorage, enabling session survival across page reloads. However, a persistent filesystem across sessions and cloud deployment support are listed as future roadmap items, indicating current limitations in long-term state management and deployment flexibility.