openclaw

Mission control for agent runs.

18
7
85% credibility
Found May 24, 2026 at 21 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Crabfleet is a team dashboard for managing AI coding assistants (called Codex agents). It lets you see everyone's active AI sessions in one place, create task cards that the AI works on, watch the AI code in real-time terminals, and review the changes when it's done. Think of it as mission control for your team's AI helpers.

How It Works

1
👋 You hear about Crabfleet

Your team mentions they have a dashboard where everyone can see all the AI coding assistants working at once.

2
🔐 You sign in

You connect your GitHub account or use a special login link to get access to your team's workspace.

3
🖥️ You see the fleet dashboard

A colorful grid shows every AI assistant session, grouped by the person who started it, with live status updates.

4
📋 You create a task card

You type a prompt like 'fix the failing test' and pick which project it belongs to. The card appears on your board.

5
You watch the AI work
👀
Watch live

You open a terminal view and see the AI typing, running commands, and making changes right before your eyes.

🤝
Take control

If the AI gets stuck, you can jump into its terminal and help out, then hand control back.

6
The AI finishes and shows its work

When done, the card moves to review and you can see exactly what files changed and what the AI wrote.

🎉 You review and merge

You check the changes, move the card to done, and the AI's work becomes part of your project.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 21 to 18 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is crabfleet?

Crabfleet is a fleet dashboard for managing AI agent runs, specifically built for teams running Codex or similar coding agents. It gives you a Kanban board to track tasks through Todo, Running, Human Review, and Done lanes, plus a live view of every cloud workspace grouped by operator. The stack is TypeScript running on Cloudflare Workers with D1 for persistence, and it ships with a Go CLI that lets you create, list, and attach to crabboxes over SSH. You can also search GitHub issues and PRs directly from the UI and turn them into cards with a keystroke.

Why is it gaining traction?

The SSH-first approach is the hook. Instead of another web dashboard, you get a CLI that feels like normal infrastructure tooling: `ssh crabd.sh new --repo owner/repo "fix the login bug"`. The fleet view grouped by person gives teams visibility into what everyone is running, which is rare in agent tooling. Multi-runtime support (Container vs Crabbox) with policy-driven selection means you can enforce repo-level defaults without manual configuration. The WebVNC and Ghostty WASM terminal grid lets you watch or take over runs in a fullscreen terminal without leaving the browser.

Who should use this?

Teams running Codex or OpenClaw agents in production who need operational visibility. DevOps leads managing a fleet of cloud workspaces across an org. Organizations with allowlist-based access control requirements that need audit trails for agent actions. If you're running one-off agent sessions without tracking who's doing what, this is overkill. If you're coordinating multiple operators across repos and need to watch, intervene, or review agent work centrally, this fits.

Verdict

This is a well-architected internal tool from OpenClaw with solid docs and a clear purpose, but the credibility score of 0.85% and 18 stars reflect its early stage and narrow audience. External contributions are not accepted, so you're adopting it as-is. The TypeScript/Cloudflare Workers stack is production-grade for their use case, but the operational complexity (SSH gateways, allowlists, runtime adapters) means a real commitment to evaluate. Worth watching if you manage agent fleets; too specialized for general adoption today.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.