crabbox  by openclaw

Remote execution and testing control plane for maintainers and AI agents

Created 1 month ago
732 stars

Top 46.4% on SourcePulse

GitHubView on GitHub
1 Expert Loves This Project
Project Summary

Summary

Crabbox provides a remote software testing and execution control plane for maintainers and AI agents. It allows users to lease cloud compute, connect to SSH hosts, or use agent sandboxes to sync code, run commands, stream output, and collect evidence, enabling a local edit-save-run loop with cloud-grade compute and agent-ready observability.

How It Works

The system consists of a Go CLI, an optional coordinator, and a remote runner. The CLI syncs code (rsync, fingerprint skip) and executes commands remotely, streaming output. The coordinator, deployable on Cloudflare Workers or Node.js/PostgreSQL, manages provider credentials, lease state, and cost controls. The data plane (SSH, rsync, execution) is direct CLI-to-runner, while the coordinator handles lease lifecycle and observability. This architecture offers flexible deployment and robust control.

Quick Start & Requirements

Install via Homebrew: brew install openclaw/tap/crabbox, or download GoReleaser archives. Prerequisites include git, ssh, ssh-keygen, rsync, and curl. Refer to "Getting started" and "Infrastructure" docs for setup.

Highlighted Details

  • Supports one-shot or warm workspaces (warmup, prewarm, run).
  • Enables named repository jobs defined in .crabbox.yaml.
  • Features efficient local-first workspace sync.
  • Provides comprehensive run observability (attach, events, logs).
  • Offers GitHub Actions hydration for consistent environments.
  • Includes failure capsule creation/replay.
  • Supports VM/workspace state checkpointing.
  • Enables emergent peer groups (--pond) for discovery and SSH-mesh.
  • Manages brokered cloud providers (AWS, Azure, GCP, Hetzner) with cost guardrails.
  • Facilitates interactive desktop, browser, and code-server leases.
  • Provides an authenticated web portal for lease/run views.
  • Collects agent workspace evidence (history, logs, artifacts).
  • Ensures stable timing records (--timing-json).
  • Features hardened coordinator authentication.

Maintenance & Community

CI runs comprehensive checks on every push/PR. Tagged pushes trigger releases. Infrastructure details are in docs/infrastructure.md. No specific community links were found.

Licensing & Compatibility

Licensed under the MIT license, permissive for commercial use.

Limitations & Caveats

The Node.js coordinator is new; Cloudflare is established. XCP-ng adapter provisions Linux only. Forwarded environment variables are restricted (NODE_OPTIONS, CI); secrets should not be CLI args.

Health Check
Last Commit

10 hours ago

Responsiveness

Inactive

Pull Requests (30d)
179
Issues (30d)
73
Star History
301 stars in the last 30 days

Explore Similar Projects

Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), Vasek Mlejnsky Vasek Mlejnsky(Cofounder of E2B), and
1 more.

pezzo by pezzolabs

0.2%
3k
Open-source LLMOps platform for streamlining AI workflows
Created 3 years ago
Updated 2 months ago
Feedback? Help us improve.