AI Tool Catalog

Code-to-knowledge-graph skill for Claude Code, Codex, and other AI coding tools. 63K stars, MIT, Python. Tree-sitter extraction across 25+ languages, Leiden community detection, NetworkX graphs. Clean pipeline: detect → extract → build_graph → cluster → analyze → report → export. 534 files, 233 edges, moderate coupling.

Graph-based RAG framework from EMNLP 2025 paper. 36K stars, MIT, Python. Combines knowledge graphs (NetworkX/Neo4j) with vector retrieval for better context. Multi-backend (Milvus, Qdrant, PGVector, OpenSearch, Redis, MongoDB, FAISS). 346 .py files, 805 edges, 11 models/tables. FastAPI server.

Open-source harness builder for AI coding. 22K stars, MIT, TypeScript + Bun. Makes AI coding 'deterministic and repeatable' via YAML workflow definitions. Uses Claude Agent SDK. 1110 files, 1151 edges. YAML-defined harnesses with provider plugins (Claude, OpenAI) and database adapters.

LLM evaluation and red teaming framework. 22K stars, MIT, TypeScript. Declarative YAML configs for prompt testing, agent evaluation, vulnerability scanning. CLI + CI/CD integration. Used by OpenAI and Anthropic internally. 134 dependencies, 2666 files (1934 .ts), 8435 edges — large but well-structured.

Secure cloud sandbox for running AI-generated code. 12.5K stars, Apache-2.0, Python + TypeScript SDKs. Provides isolated cloud environments (Firecracker microVMs) where agents can execute code safely. 738 files, 798 edges. SDK is clean; infrastructure is cloud-hosted.

Open-source AI gateway for routing to 1600+ LLMs with guardrails, fallbacks, load balancing, and caching. 12K stars, MIT, TypeScript. Runs on Cloudflare Workers, Node.js, or Docker. 761 files, 1879 edges. Hono framework, Redis caching, JWT auth. MCP gateway support.

Code search MCP for Claude Code, Cursor, and Codex. 11.7K stars, MIT, TypeScript. Tree-sitter parsing, Merkle-tree caching for incremental re-indexing, embedding-based semantic search via Voyage AI. Positions itself as making the entire codebase available as context for coding agents. 71 files, 94 edges, clean architecture.

Browser-based plan and diff review tool for AI coding agents. 6K stars, Apache-2.0, TypeScript. Supports Claude Code, Copilot CLI, Gemini CLI, OpenCode, Codex, and more. Inline annotations, structured feedback back to agents, encrypted sharing. Uses Claude Agent SDK and OpenAI Codex SDK.

Natural language E2E testing framework. 5.6K stars, MIT, TypeScript. Write tests in plain English, AI (Anthropic Claude) executes them via Playwright. Interesting concept but SaaS-oriented (Clerk auth, Vercel Postgres, Stripe billing). 78 dependencies, 238 files, low coupling (53 edges).

Self-hosted AI agent orchestration dashboard. 5K stars, MIT, TypeScript. Task dispatch, multi-agent workflows, cost monitoring, SQLite backend. Next.js 16 + better-sqlite3. Zero external dependencies. The CityCommand competitor we didn't know existed.

Capability-based OS-level sandbox for AI agents, built by the Sigstore team. Rust, Apache-2.0, v0.74.0. 3.8K stars, 252 forks. Endorsed by Datadog and Okta security engineers. Three-layer tool mediation: invocation policy (argv gating), credential brokering (phantom tokens), L7 endpoint policy (HTTP method+path filtering). First evaluation in WHETREV-172 addressing agent runtime security at the OS level. Rated 7/10.

Claude Code skill for generating reveal.js slide decks. MIT, 14 files, 29 stars. Pure skill shape (SKILL.md + reference markdown, no code). Preview-gate workflow: 2-slide preview, gate on approval, build full deck by appending. Density axis (spoken-over vs read-alone) cascades through type size, word count, slide count, layout. Rated 6/10 in Whetstone WHETREV-172 series.

You wrote a prompt. It works. But is it good? Simmer runs your artifact through multiple rounds of criteria-driven refinement — each round, a panel of judges reads your code, understands the problem, and proposes specific improvements.

Open-source community intelligence platform that monitors 25+ AI/tech data sources, processes signals through LLM-powered agents, and surfaces actionable market intelligence — from trending topics and product traction to talent gaps and billion-dollar market opportunities.

**A graph-first IDE for agentic coding — desktop app, or served to a browser

No summary available.

Claude Code skill for motion-first web interfaces. MIT, 70 files, 8 deployed production sites (German SMBs). Introduces the capability catalog skill shape — a curated vocabulary of techniques that expands what the agent considers, distinct from workflow skills and domain adapters. Token architecture: 9 CSS custom properties + 1 radius knob + 1 font = complete visual identity. Rated 8/10 in Whetstone WHETREV-172 series.

Data, Model, Recipe for an open Jev

No summary available.

No summary available.

No summary available.

No summary available.

A browser agent with a dynamic, indexed action space.

No summary available.

Demand/discharge ledger for Claude Code hooks with 24 clauses implementing five construction types for guard design. Repricing thesis: mandates are counterfeits because they get suspended when needed; repricing swaps which fate is cheap so forgetting produces the safe outcome. POSIX shell, 3.8K stars equivalent community interest. Rated 8/10 in Whetstone WHETREV-172 series.

AI coworkers you can hand real work to, and actually trust with the access. Each gets a computer of its own: a real browser with its own logins, its own files, and only the tools you grant. Every action decided before it happens and recorded after.

A personal agent with a browser, terminal, files, and work that keeps going. Compatible with any agent harness.

This package provides an implementation of the CL API to assist with local development of applications that can run on a Cortical Labs CL1 system.

Recurring autonomous agent scheduler (cronloop.ai, Cactal Inc). No repo — evaluated from production deployment data: 22 city-specific agents run for 25 days curating AI events for aievents.now. Five production learnings, three extractable patterns. Strongest external validation of durable memory as highest-leverage improvement for recurring agents. Rated 7/10 in Whetstone WHETREV-172 series.

most recent version downloadable in https://adame.ai

No summary available.

Context aware, pluggable and customizable PII de-identification service for text and images.

No summary available.

An open benchmark and harnesses for AI in cybersecurity operations.

A small code-review workflow built with TypeSafe Jev. It can review a Git diff or scan a complete codebase, follows the strongest structured signals through focused model calls, and presents the result in a quiet local dashboard.

No summary available.

CPU/GPU Converter from E-Book to audiobook with chapters and metadata

A modern analytics pipeline for tracking and analyzing GitHub contributions across repositories. The system processes contributor data, generates AI-powered summaries, and maintains a leaderboard of developer activity.

~2 million current job postings, crawled daily from ~65,000 company career sites (Greenhouse,

No summary available.

RAG matches chunks. GraphRAG structures context. M-flow scores evidence paths.

Build trusted agent skills. Govern their marketplace. Improve what your organization knows.

No summary available.

No summary available.

No summary available.

No summary available.

English｜简体中文｜繁體中文｜[日本語](/translations/README.ja

No summary available.

Prose editing pipeline + anti-sycophancy skills (humanist + candor by itsryanlenk). Backed by ~1,300 blind judgments across two study designs showing quality-detection dissociation: processed AI prose beats human 18/18 on quality while detected 97.8% of the time. Detection evasion is structurally impossible. Discrimination ratio method for evaluating any rule's diagnostic value. false-positive CI gate with 24 known-human documents. Rated 9/10 — highest in Whetstone WHETREV-172 series.

Small Jev-like decision models you can train and run yourself.

让每一个 AI 创业者都拥有自己的商业闭环

[](https://github.com/kenryu42/cc-safety-net/actions/workflows/ci.y

No summary available.

English | 中文 | 日本語

No summary available.

Universal memory layer for AI agents. 58K stars, Apache-2.0, Python. Supports 15+ vector stores, 10+ LLM providers, entity extraction via spaCy. SaaS + self-hosted. Very broad — tries to be everything to everyone.

MemOS 2.0 Stardust（星尘）

🤗 Models on Hugging Face | Blog | Website |

🌍 English | 日本語 | 简体中文 | 한국어

No summary available.

No summary available.

No summary available.

👉 Microsoft Research Blog Post

MagenticLite is the next generation of Magentic-UI — an agentic application from Microsoft AI Frontiers, redesigned to do more with less. It pairs an on-device-friendly orchestrator model

The Python Risk Identification Tool for generative AI (PyRIT) is an open source

No summary available.

Open-weight typed decisions on Apple Silicon. Core ML, Neural Engine, zero generated tokens.

Open-weight typed decisions, running natively on Apple Silicon.

No summary available.

16 portable editorial systems encoded as Claude Code Agent Skills. Key structural innovation: 'Protect the {domain}' opening section is a genuine construction (type 2: one door) forcing preservation before rewriting. Testable avoid lists are type 3 (resident failing case). Rated 7.5/10 in Whetstone WHETREV-172 series. our agent dispatch system stylebook integration prototyped from this evaluation.

No summary available.

🔥 Open Design 0.13.0 — Stay in Flow is here. Long design sessions used to break on every interruption — a run lost its place, a model picker made you guess, an export needed one more detour. 0.13.0 keeps the session alive: resume Codex / OpenCode / Pi / Open Design Cloud runs across turns, p

Current version: 0.2.0

__ _ _

No summary available.

No summary available.

The IDE for Your Coding Agent

No summary available.

CFNext —— 部署在 Cloudflare Workers / Pages 的代理订阅管理面板，单文件同时实现 VLESS / Trojan / XHTTP 多协议代理与图形化配置：/UUID（或自定义路径）进入面板，/sub 输出订阅；内置节点优选器、节点数量控制、轮询下发、ECH 加密、落地与出站代理及地区 / 运营商 / IP 版本多维筛选，部署即用，绑定 KV 后改完即生效。

A pointing tool for working with AI agents. It's "screenshot" crossed with "mugshot", taken by someone who already knows they're right.

No summary available.

No summary available.

No summary available.

No summary available.

Simple and flexible library for Reservoir Computing architectures like Echo State Networks (ESN).

RSS-Bridge is a PHP web application.

A Powerful Few-shot Voice Conversion and Text-to-Speech WebUI.

No summary available.

Write design system rules that agents can verify.

No summary available.

No summary available.

No summary available.

A verified game-dev library · live engine tools · project memory. One npm package.

No summary available.

Claude Code plugin that replaces the compaction summary with Jev decisions:

Open-source apps that replace a SaaS product, running in your own Cloudflare account.

Can we run something like Jev on a 3090 at home?

Semantic ifs from open models, on a 3090 at home.

No summary available.

No summary available.

English | 简体中文

No summary available.

Cognee - The Open-Source AI Memory Platform for Agents

Website | Docs | Issues | Example projects | Feature requests | [Public roadmap](https://triggerdev.featurebas

Join our Discord community for support, discussions, and updates:

Easy, fast, and cheap LLM serving for everyone

ModelTrace 是一个本地运行的主动模型归因工具。它通过三条独立的长整数生成挑战提取输出指纹，在统一候选库中自动判断模型家族和具体版本。

Coding-agent-driven creation of editable, tastefully crafted visual artifacts.

版本：0.2.0 | 更新：2026-09-17

The ultimate resource hub for AI-powered game development tools. Discover cutting-edge LLMs, World Model, Agent, Code, Image, Texture, Shader, 3D Model, Animation, Video, Audio, Music, Singing Voice and Analytics. 🔥

The First AI-Native Game Engine. 💥

简体中文 | English

No summary available.