WeChat (Linco Lab)
ai-toolsai-engineering
Original source

本周GitHub爆火项目:值得收藏AI Agent工具

Summary

Linco Lab’s weekly GitHub Trending roundup (data as of 2026.06.28), featuring 6 trending AI Agent projects.

  1. Agents taking over long tasks — video production, research, coding shifting from Q&A to executable workflows
  2. Memory + MCP heating up — codebase memory, long-term memory, knowledge graphs becoming foundational Agent infra
  3. Specifications becoming AI-readable — design systems, skills, security frameworks as structured files for Agents

The 6 Projects

#ProjectStars (weekly)What It Does
1calesthio/OpenMontage25,336 (+18K)Open-source Agent video factory — Agents orchestrate scripts, materials, editing; 52 tools/skills
2DeusData/codebase-memory-mcp17,728 (+7.7K)Persistent code knowledge graph via MCP for coding agents; single binary, zero deps
3google-labs-code/design.md22,362 (+6K)DESIGN.md spec for visual identity so coding agents maintain consistent brand/design
4Panniantong/Agent-Reach43,503 (+7.7K)Multi-platform search CLI for Agents (Twitter, Reddit, YouTube, GitHub, Bilibili, Xiaohongshu)
5bytedance/deer-flow75,056 (+3.3K)Long-task SuperAgent framework — research, coding, creation for minute-to-hour tasks; sandboxes, memory, subagents
6topoteretes/cognee24,024 (+5.9K)Self-hosted Agent long-term memory via knowledge graphs; links docs, projects, user prefs across sessions

Takeaways

  • Agents gaining real work capabilities (video, search, code, design) — not just chat
  • Context (memory/MCP/design specs) becoming the core asset of AI products
  • Real value: writing personal workflows as specifications, skills, and reusable tool chains
  • Key question: whether Agents can move from demos to stable delivery