Trending AI agent repos, explained properly.
Every week we track the repositories the internet is starring — then write up what each one actually does, who it’s for, when to skip it, and ten automations a real team could build with it.
Editor's picks
Superpowers
Stops your coding agent jumping straight to code — a full development methodology delivered as composable skills.
Top Trending Github repos this week
- 01Prime AgentSelf-improving RLM agent for long-running coding and research tasks.★ 10.7k
- 02HoleheCheck if an email is registered on 120+ sites.★ 6.7k
- 03SemanticaGraph-native context for accountable AI.★ 5.1k
- 04TencentDB Agent MemoryTeam agent memory hub — Chat Memory, Skills, and LLM-Wiki shared across agents.★ 4.4k
- 05Agent SkillsProduction engineering skills for 70+ coding agents.★ 3.8k
- 06Cloudflare ComputerDurable agent filesystem and isolated exec environment inside Cloudflare Workers.★ 2.9k
- 07Google Agent SkillsOfficial Agent Skills for Google Cloud, Ads, Analytics, GKE, and BigQuery.★ 2.2k
- 08Code Graph RAGMonorepo code RAG via Tree-sitter for knowledge graph queries and edits.★ 1.7k
- 09CelldSelf-hosted distributed Durable Objects using S3/GCS as the only control plane.★ 1.5k
- 10SwitchyardRoute LLM traffic and translate OpenAI ↔ Anthropic payloads natively.★ 1.2k
Browse by what it does
- Agent frameworks21Scaffolding for building agents that plan, call tools, and run multi-step work.
- Agent skills17Reusable instruction packs that make a coding agent better at one specific job.
- Coding agents13Terminal and desktop agents that read, write, and ship code for you.
- Workflow automation11Visual or code-first platforms for wiring triggers, steps, and integrations.
- Memory & context9Storage layers that let an agent remember across sessions and teammates.
- RAG & knowledge12Turn your docs, code, and databases into something an agent can answer from.
- Browser & data capture6Drive a real browser or scrape structured data for lead-gen and research.
- LLM gateways4One endpoint in front of many model providers — routing, fallback, cost control.
- Observability & evals2See what your agents did, what it cost, and whether the output was any good.
- Vector & search2Databases and engines that make semantic retrieval fast at production scale.
- Vertical agents5Agents purpose-built for one business function — ads, finance, hiring, content.
- Learning6Courses, curricula, and reference collections for getting your team fluent.
The whole catalog
6 repositories
- Learning132kShubhamsaboo
Awesome LLM Apps
100+ working agents, skills, and RAG apps you can clone, ship, or sell — Apache 2.0, tested end to end.
PythonPlug in - Learning117kmicrosoft
Generative AI for Beginners
Microsoft's 21-lesson course on actually building with generative AI — the applied companion to their fundamentals curriculum.
Jupyter NotebookPlug in - Agent frameworks92kpunkpeye
Awesome MCP Servers
The index of MCP servers — check here before you build an integration that already exists.
Plug in - RAG & knowledge78.6kEgonex-AI
Understand Anything
You just joined a team with a 200,000-line codebase. This turns it into a graph you can explore and question.
TypeScriptPlug in - Learning73.7kshareAI-lab
Learn Claude Code
Builds a working agent harness from zero so you understand what these tools are actually doing.
PythonPlug in - Learning64.2kmicrosoft
AI for Beginners
Microsoft's free 12-week, 24-lesson AI curriculum — the least glamorous and most reliable way to get a team fluent.
Jupyter NotebookPlug in
What it actually does
Plain English, no marketing voice. If a repo's README is three languages of badges, we read it so you don't have to.
When to skip it
Every entry says when the tool is wrong for you. That section is usually more useful than the one above it.
Ten automation ideas
Tied to workflows a real team runs by hand — marketing, sales, support, finance, ops. Ideas, not tutorials.
Reading about automation is not the same as having it.
Every repo here is a building block. LimeDock assembles them into the workflows your marketing, sales, and ops teams run by hand every week — deployed into your own cloud, wired into your Slack and CRM, and owned by you.
Book a workflow call