Data sources: GitHub Trending dump issues Rain1er/trending #393 (Oct 2 daily, updated Oct 3 00:32) and happydog-intj/ai-xiaohongshu-daily #527 (‘2026-10-03’, created Oct 2 UTC), compared with the prior day’s #524. Primary metric: stars today. Repos above 40k total stars (Agent-Reach, caveman, superpowers, ponytail, impeccable, mattpocock/skills, marketingskills, hyperframes, codegraph) are excluded. NVIDIA/OpenShell (+584, down 77% from +2,503) and mksglu/context-mode (+276, down 23% from +357) are excluded for slowing growth. pablostanley/yoinks (+629) is not AI-related (a video downloader); cursor/plugins (+168) and google/skills (+78) had gains too small to qualify. One repo passed today’s criteria.
mvschwarz/openrig — +691 ⭐ today
https://github.com/mvschwarz/openrig
Total stars: 4,121 · Daily gain: +691 (+8% vs. the previous day’s +640; -6% vs. its last featured day, Sep 30, at +733) · Language: TypeScript · License: Apache-2.0
OpenRig turns separate AI coding agent sessions such as Claude Code, Codex and Pi into a coordinated, persistent team with defined roles and shared context. You declare a team topology (pods, edges, continuity policies) in RigSpec YAML; a local daemon (a Hono HTTP server) launches each agent on top of tmux, stores state, topology and telemetry in SQLite, and exposes the team to both humans and agents through a CLI/TUI and an MCP server. Its key difference is that it only owns the coordination layer and leaves agent execution alone, so you can drop into any agent’s terminal directly and mix agents from different providers in one team. Main features:
- Declarative team topology via RigSpec YAML
- Persistent seats with stable addresses (e.g. dev-owner@first-project) and owned work queues
- Full team snapshot and restore, plus discovery and adoption of existing tmux sessions
- Cross-agent messaging with rig send, rig broadcast and rig chatroom
- Per-seat typing guards and sandbox modes, Slack integration, and context-usage telemetry
After its first surge in late September and three days of cooling, it rebounded today (+8% day over day). Demand for running several coding agents in parallel as a ‘team’ keeps growing, and OpenRig’s reboot-proof teams and handoff workflows appear to be drawing renewed attention.
Practical takeaway: Pilot an intake → planning → implementation → review pipeline across different agents locally, with a clear audit trail of who owns what.
Tags: #MultiAgentOrchestration #ClaudeCode #CodingAgents #tmux #AgentTeams #MCP
Re-listing note: openrig is within a similar range of its last featured day (Sep 30, -6%) and rebounded 8% day over day, so it returns to the radar. This is not a consecutive streak, since it was not featured on Oct 1 or Oct 2.
Sources: Rain1er/trending #393, happydog-intj #527, happydog-intj #524
Leave a comment