My curated AI indexes on GitHub: what each list covers

I keep a lot of lists. Over time the lists I make while looking into an area of AI have turned into repositories. Fifteen of them are about AI now, and this post is the guide to them: what each one covers, roughly how big it is, when it was last touched, and which ones are frozen snapshots rather than living documents. The map of everything, not just the AI part, is Start-Here, which is regenerated from the GitHub API and is the place to begin if you want to find your way around my repos.

A caveat that applies to all of them: these are indexes of other people's projects. I list what I found, I mostly haven't run it, and inclusion is not endorsement. Where a list is a point-in-time snapshot I say so below, because in this field a six-month-old list is already a historical document. (One related list, the Obsidian one, has its own post and its repo, so I've left it out here.) None of the fifteen was empty or trivial enough to drop; the smallest, the harnesses list, is short but covers a term that is still being defined.

Agents and harnesses

Local-AI-Agent-Resources is the biggest of the agent lists: about 130 projects in 17 sections, covering agents whose control loop runs on your own machine. The model can still be a hosted API; what matters is that the harness is local. Sections run from CLI coding agents and tooling around them, through computer-use agents, desktop GUIs, personal assistants, memory, gateways and protocols, to voice agents and research assistants. It has a scope file with inclusion rules. Last updated 6 April 2026, and it says it is refreshed on a rolling basis.

AI-Harnesses is much smaller: 26 projects that describe themselves as agent harnesses, plus a short list of guides. "Harness" only became a common term in late 2025 and early 2026, and the README opens with a note on how I'm using it (the layer that runs the agent: tool dispatch, permissions, context, lifecycle, as opposed to a framework for building one). It is explicitly a snapshot from April 2026.

Awesome-AI-Agents-And-Assistants is the oldest and, by now, the most historical: an ecosystem map of agent and assistant platforms as they stood in February 2025, split into category pages (development frameworks, development tools, agent types such as computer-use and voice agents) with a separate file of my opinionated notes. Last pushed March 2025. Read it as a record of what the landscape looked like then.

Agents-Index is a different kind of index: it lists my own work, the 80-plus AI agent configurations and system prompts I've shared on GitHub, with interfaces and a pointer to other indexes. It says it is updated periodically; the last update was October 2026 (pushed 30 October).

Coding tools and Claude Code

Awesome-AI-Coding-Tools is a snapshot generated on 1 September 2025, and it says so at the top. It is a deliberately partial map of the AI coding space, about 110 repositories grouped under headings like AI IDEs, general CLIs, agent unification, code review, code completion, context tools, and Claude Code accessories. The long tail of coding CLIs is enormous, so the aim was the shape of the field rather than completeness. The repo was touched again in August 2026, but the content is still that September 2025 snapshot.

Cool-Claude-Code-Stuff is the largest list I have by raw count: several hundred entries (my count of the headings is around 370) under alphabetical categories from account management and ACP through hooks, harnesses, MCP, monitoring and so on. It is dated 2 October 2025 and hasn't been refreshed since, so anything newer than that, including the plugin ecosystem as it is now, isn't in it. For my own Claude Code plugins there's a separate post.

Evaluations and LLM councils

Awesome-AI-Evaluations-Tools is the one I'd call alive. It lists open-source tools, frameworks and benchmarks for evaluating AI systems (about 130 entries), with sections for platforms, benchmarks, agentic and tool-use evals, browser and web-agent evals, coding, multimodal and vision, RAG, hallucination and grounding, bias, sycophancy, and refusal and safety. It has a flat CSV of every entry (category, stars, last commit, description) regenerated weekly by a GitHub Action, so the data file stays fresher than the prose. Last pushed 2 November 2026.

Awesome-LLM-Council-Projects indexes projects built on Andrej Karpathy's llm-council idea, where several models deliberate and a final model synthesises an answer. About 60 projects, grouped by what they are: benchmarks, plugins, MCP servers, SDKs, skills, mobile and IDE versions, local setups, and a section for my own variations. It describes itself as a non-exhaustive snapshot from April 2026 (pushed 23 April).

Simulation, society and geopolitics

AI-Synthetic-Society-Experiments collects projects that use many interacting agents as a stand-in for real social dynamics: foundational research such as Stanford's generative agents, civilisation simulations, virtual towns, emergent behaviour, political modelling, and the infrastructure underneath. About 35 projects in 11 sections. The scope note excludes general multi-agent frameworks that aren't about simulating society. Last pushed 5 April 2026.

AI-Geopol-Projects is the narrower, geopolitical cousin: roughly 20 projects on wargaming, policy simulation, conflict modelling and strategic forecasting with LLM agents, including a few of mine (Geopol Modeller, Geopol Forecaster and Geopol Forecast Council). Last updated 19 April 2026.

Monitoring-The-Situation is the odd one out, and arguably the most practical. It's an index of open-source pieces for building a command-centre style news dashboard: situational-awareness and geopolitical dashboards, news grids, RSS backends and deduplication, OSINT and information-fusion platforms, digital signage, kiosk management, and AI analysis layers. Around 70 projects over about 20 sections. The scope is narrow on purpose and excludes server-metrics dashboards. April 2026 (pushed 21 April).

Personal AI

Personal-AI-Resources covers memory and persistence, context engineering, personal assistants, second-brain tools, personal agents and automation, personalisation, email, and data management: about 110 projects. A flaw you should know about: the README's badge says April 2026 while its note says last updated April 2025, and the repo was pushed in April 2026, so treat it as an April 2026 snapshot and the note as a typo I haven't fixed.

Voice and speech

Awesome-Whisper-Apps lists around 85 apps and libraries built on OpenAI's Whisper, with a quick-start guide by platform and use case (voice typing on Linux, Mac or Windows, subtitles, meeting transcription, self-hosted web UIs, mobile). Last pushed 13 August 2026.

Linux-Voice-Tech is wider than Whisper and narrower in platform: voice technology you can actually use on Linux. Its own badge says over 120 resources, and it includes a Wayland-specific section (dictation tools that can type into Wayland windows), plus GUI and CLI voice typing, streaming STT, self-hosted servers, cloud APIs, voice assistants, TTS and MCP servers. It has a notes file on inclusion criteria and a getting-started page. The badge says last updated 29 March 2026; the repo was pushed again on 19 October 2026.

How I keep them, and how to use them

The honest summary of maintenance is that two or three of these are living (evaluations, voice, the agent index) and the rest are snapshots with a date in the first lines. I'd rather date a list than pretend it is current. Star counts and last-commit dates on the entries are mostly shields.io badges, which update on their own, so the entry-level freshness is better than the list-level freshness. If you want to know what has changed since a snapshot, those badges are the quickest signal.

If you only open one, pick the one that matches the thing you're about to build or choose. And if none of them matches, Start-Here links the full set of my topic indexes, which is about 64 hand-curated index repos as of its last rebuild on 2 November 2026.