runtm  by runtm-ai

AI coding agents build and deploy software in isolated sandboxes

Created 5 months ago
272 stars

Top 94.7% on SourcePulse

GitHubView on GitHub
1 Expert Loves This Project
Project Summary

Summary

Runtm provides open-source, OS-level sandboxed environments designed for AI coding agents to build, test, and deploy software. It enables agents like Claude Code and Cursor to operate with full permissions in isolated workspaces, mitigating risks to host systems, and instantly deploys their creations to live, shareable HTTPS endpoints with integrated logging and previews.

How It Works

The core of Runtm utilizes OS-level sandboxing technologies, specifically bubblewrap on Linux and seatbelt on macOS, to create highly isolated, ephemeral environments. AI agents interact with these sandboxes, which offer instant startup times (<100ms) without the overhead of traditional containers. Once an agent completes its work, a single command deploys the application to a real URL, providing live access, logs, and previews for immediate iteration and debugging.

Quick Start & Requirements

  • Installation: Install the Python CLI via uv tool install runtm or pipx install runtm. A Go CLI (runtm-api) is also available via a curl command.
  • Prerequisites: Requires Linux (for bubblewrap) or macOS (for seatbelt) for OS-level sandboxing. Python 3.x is needed for the Python CLI.
  • Links: Full documentation is available at docs.runtm.com.
  • Setup Time: Sandboxes start in under 100ms.

Highlighted Details

  • OS-level isolation using bubblewrap (Linux) or seatbelt (macOS).
  • Sandboxes start in <100ms, eliminating container overhead.
  • Supports multiple AI coding agents including Claude Code, Codex, Gemini CLI, and Cursor.
  • Persistent workspaces allow sessions to be stopped and resumed with files preserved.
  • Deployments are accessible via live HTTPS endpoints with integrated logs and previews.

Maintenance & Community

Support is available by opening an issue on GitHub or joining the project's Discord community.

Licensing & Compatibility

  • Server components (API, worker, infra) are licensed under AGPLv3.
  • CLI, Sandbox, and Shared components are licensed under Apache-2.0.
  • Templates are licensed under MIT.
  • The AGPLv3 license for server components imposes copyleft requirements, meaning derivative works of these components must also be made available under AGPLv3.

Limitations & Caveats

The OS-level sandboxing features are dependent on Linux or macOS environments; Windows support for sandboxing is not explicitly detailed. The AGPLv3 license for core server components may present compatibility challenges for integration into closed-source commercial products.

Health Check
Last Commit

2 days ago

Responsiveness

Inactive

Pull Requests (30d)
4
Issues (30d)
5
Star History
154 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.