Collection basis: GitHub Trending bot issue happydog-intj/ai-xiaohongshu-daily #593 (“GitHub Trending AI 早报 · 2026-10-25”, 00:00 slot, Top 8), compared with the previous day’s #590 (Oct 24, Top 6). Figures come from the bot issue, and stars today is the primary metric. Excluded: repos with 40k+ total stars (cloudflare/security-audit-skill 43.6k, claude-howto 44.3k, impeccable 86.1k, gemini-cli 107.8k); slowing momentum versus the last featured day (NVIDIA/OpenShell +1,240 → +893, −28%); low growth (aws/agent-toolkit-for-aws +11); and a July-launched repo re-trending at a low growth rate (Nutlope/hallmark +567, 1.8% of total).
makecindy/cindy — +471 ⭐ today
https://github.com/makecindy/cindy
Total stars: 3,895 · vs. previous day: not listed on Oct 24 → +471 (new entry, 12.1% of total) · Language: TypeScript · License: Apache-2.0
Cindy is an open-source AI agent that works out of the box under the tagline “Consider it done,” combining multiple harnesses, models and tools into one agent that does real work on your local files and logged-in apps. Its first supported harnesses are Claude Code and Codex, with a native harness in development, and you can switch models or harnesses mid-task while the workspace, memory, skills and tools carry over. The codebase is a pnpm monorepo with an Electron desktop app (apps/desktop), an Expo/React Native mobile app (apps/mobile), shared packages for auth, device linking, agent orchestration and model providers, and per-platform bundled Claude Code, Codex and ripgrep binaries; its “Orca” multi-agent orchestration lets agents on different harness/model combinations plan, execute in parallel and review a single task. Key features:
- Drives your browser, computer and phone, and takes work from IM and schedules
- Persistent memory: corrections stick and are shared across harnesses
- Reusable skills: teach a workflow once and reuse it everywhere (team sharing in development)
- Automation: schedules recurring work, runs it and reports back
- MCP connections to internal tools and business systems, plus plugins (via SkillHub or manual install)
- Model access: the official Cindy service, existing Claude Code/Codex coding plans with no double billing, your own API keys, or local models
Its differentiator is turning existing coding-agent subscriptions into a general-purpose work agent. After collecting roughly 2,000 stars in its first three weeks, it just broke into the GitHub Trending AI top ranks for the first time (no GitHub Trending #1 or day/week/month #1 badges on Trendshift; best Trendshift Top 10 rank of #10 on Oct 24).
Practical takeaway: Teams already paying for Claude Code or Codex can trial desktop/mobile document and app automation plus scheduled tasks at no extra model cost (official builds include TapDB usage analytics, so consider building from source for enterprise use).
Tags: #AIAgents #MultiAgent #ClaudeCode #Codex #DesktopAgent #AgentOrchestration
Leave a Reply