Twelve community-maintained "awesome lists" on GitHub keep the AI-coding tool catalog current — and in 2026 they update faster than any single vendor's docs. awesome-chatgpt-prompts (Fatih Kadir Akın) has passed 100,000 GitHub stars and added Claude- and Cursor-specific prompts since 2024; awesome-mcp-servers (Frank Fiegel of Glama) went from a side project at the November 2024 Model Context Protocol launch to the de-facto index of MCP servers in under a year. The lists below are the ones still being updated, grouped by what they actually catalog: tools, prompt collections, frameworks, and learning resources.
Why awesome lists, not vendor docs?
Vendor docs cover one product. An awesome list catalogs the field. When OpenAI open-sourced the Codex CLI in April 2025 and Anthropic shipped its Claude Code MCP server ecosystem the same year, the canonical catalog of "all the agentic coding tools right now" was not on either company's site — it was awesome-ai-agents on GitHub, maintained by the E2B team and dozens of outside contributors. That is the structural advantage: a list maintained by people with no vendor stake adds new entrants the day they launch and removes ones that died, while vendor pages still link to competitors that shut down 18 months ago.
The flip side: awesome lists rot too. Spend an afternoon clicking through entries on any list older than two years and you will hit dead links, archived repos, and tools that pivoted to a different niche. Star count is a weak proxy for freshness — check the last commit date and how many open PRs are sitting in the queue before trusting an entry.
Tools — agents and AI-coding products
awesome-llm-apps — Shubham Saboo
End-to-end LLM application examples with source code — RAG, agents, multi-agent teams, and voice apps. Strong for developers who want to read working code rather than another tutorial.
awesome-ai-agents — E2B team
Open-source and commercial agent frameworks side-by-side: AutoGPT, BabyAGI, OpenDevin, MetaGPT, plus the agentic-coding tools (Aider, Devin, Cursor agent mode). Maintained by E2B, which builds the sandbox most of these run inside.
Awesome-LLM — Hannibal046
The canonical landscape index — papers, models, fine-tuning frameworks, deployment tools, and benchmarks. Less coding-specific than the others on this page, but it is the list every other AI/LLM list ends up citing, so it doubles as a sanity check when you want to know whether a project is established or noise.
Prompt collections — rules, system prompts, skills
Prompt repos are where tribal knowledge lives. Cursor rules and Claude Code skills each have their own canonical lists, and the leaked-system-prompts repo is how teams reverse-engineer what the commercial agents are actually being told.
awesome-chatgpt-prompts — Fatih Kadir Akın (f)
The most-starred prompt repo on GitHub. Started for ChatGPT in 2022, now covers Claude, Cursor, and other assistants with role-based prompt patterns ("act as a senior reviewer", "act as a SQL tutor").
awesome-cursorrules — PatrickJS
Rules files for Cursor's .cursor/rules/*.mdc system, organized by stack: Next.js, Astro, Rails, Django, Tauri, and dozens more. The legacy.cursorrules single-file format is still here for older installations.
awesome-claude-code — hesreallyhim
Slash commands, hooks, sub-agents, and CLAUDE.md examples for Claude Code. The list to bookmark if you are building a custom harness on top of Claude Code rather than using it out of the box.
awesome-copilot — GitHub (Microsoft)
Microsoft-maintained collection of GitHub Copilot custom instructions, prompts, and chat modes. Worth checking for the official examples even if you are not on Copilot — the patterns transfer to Cursor and Claude Code with minimal edits.
leaked-system-prompts — jujumilk3
Public archive of leaked system prompts from Cursor, v0, Bolt, Devin, Cline, Claude.ai, and many others. Reading them is the fastest way to understand how commercial agents structure tool calls and refusal behavior — see also system prompts for agentic coding.
Frameworks — orchestration, MCP, agent libraries
awesome-mcp-servers — Frank Fiegel (Glama)
Hundreds of Model Context Protocol servers, grouped by category: databases, browsers, design tools, ticket systems, observability. The first place to look before writing your own MCP server — there is a good chance it already exists.
awesome-langchain — Kyrolabs
LangChain ecosystem — chains, agents, memory backends, tools, and adjacent libraries (LangGraph, LangSmith, LlamaIndex bridges). Useful even if you are not using LangChain itself, because most adjacent agent frameworks integrate with these primitives.
Learning resources — courses and curated guides
awesome-generative-ai-guide — Aishwarya Naresh Reganti
Structured learning roadmap with weekly papers, interview questions, and hands-on notebooks. Stronger on the research side than the agent-coding side, but it is the most coherent self-study path of the lists here.
llm-course — Maxime Labonne
Three-track course (Fundamentals, Scientist, Engineer) with Colab notebooks for fine-tuning, quantization, and RAG. Engineer track is the directly relevant one for AI-coding tooling work.
How do I tell which list is still maintained?
- •Last commit date. Open the repo, sort by recent commits — if nothing has landed in six months, treat the list as a snapshot and verify each entry separately.
- •Open-PR count. A healthy list has 5–30 open PRs from outsiders adding entries. Zero open PRs usually means the maintainer stopped reviewing.
- •Sample three random entries. Click them. If two of three are archived, 404, or pivoted to a different product, the list is stale regardless of star count.
- •Watch (Custom → Releases). Better than starring — you get a notification when the maintainer cuts a new tagged version, which is a stronger signal than a casual edit.
Where the conversation actually happens
Awesome lists catalog things; communities discuss them. New entries usually get debated on Discord and Reddit before they land in a list — see AI coding communities for the active forums, and AI editor comparison when the list entries blur together and you need a side-by-side. For workflow-level guidance, agentic coding workflows is the operational counterpart to the tool catalogs above, and agentic coding levels is the autonomy framework that decides which tier of tool you actually need.