xClean Tools

AI Repos — Daily Top 5

UPDATED 2026-11-02 13:14 PST

2026-11-02 · MONDAY · 13:14 PST

  1. 1 vercel-labs/agent-desktop +2,243 STARS TODAY · ★ 9,236 · RUST Vercel Labs opened agent-desktop, a Rust command-line tool that lets AI agents drive a desktop computer. It works through operating-system accessibility APIs on macOS and Windows and ships an MCP server, agent skills, probes and benchmarks in the repo. It added 2,243 stars today, the largest single-day gain in the pool.
  2. 2 p-e-w/heretic +404 STARS TODAY · ★ 38,009 · PYTHON heretic modifies an open-weight language model's behavior without retraining it, using abliteration to strip refusals and to damp down slop and sycophancy. The Python repo ships one preset config per behavior plus a classifier and a PIQA config for checking that capability survives. It has 38,009 stars, 404 of them added today.
  3. 3 mvanhorn/last30days-skill +141 STARS TODAY · ★ 66,358 · PYTHON last30days-skill is an agent skill that researches a topic across Reddit, X, YouTube, Hacker News, Polymarket and the web, then returns a summary tied back to its sources. The repo carries plugin manifests for Claude Code, Codex and Grok harnesses over Python and Go collectors, with READMEs in seven languages. It stands at 66,358 stars.
  4. 4 max-sixty/worktrunk +136 STARS TODAY · ★ 9,969 · RUST worktrunk is a Rust command-line tool for Git worktree management, written for running several coding agents side by side. The repo includes Claude Code and Codex plugin directories, skills, hooks, benchmarks and a Nix flake. It has 9,969 stars, 136 added today, and was last pushed this morning.
  5. 5 RabbittuX/purrge ★ 227 · PYTHON purrge is a macOS disk cleanup tool that hands the judgment call to a coding agent. It warns when space runs low, has Claude Code or Codex rate each candidate file or cache, and deletes only what the user ticks. The Python repo ships install scripts, a packaging directory and tests, and has 227 stars since it was created on October 20.

2026-11-01 · SUNDAY · 13:12 PST

  1. 1 ifixai-ai/iFixAi +1,654 STARS TODAY · ★ 42,419 · PYTHON iFixAi, a Python tool for independently auditing AI agents, added about 1,654 stars today to reach roughly 42,400. It is built to answer whether an agent is doing what it is supposed to do, run either by a human or by the agent itself, and the project claims a verdict in under 120 seconds. Its listed targets include the EU AI Act, ISO 42001 and the OWASP LLM risks.
  2. 2 scenario-labs/skills +98 STARS TODAY · ★ 1,447 · PYTHON Scenario Labs' set of agent skills for generative media production added 98 stars to reach about 1,450. The skills pick a model for images, video, audio or 3D, price a job before spending, and hold characters and brands consistent through the Scenario MCP server. Separate skill sets drive Blender, Maya, ZBrush, Unreal and Unity.
  3. 3 kornia/kornia +88 STARS TODAY · ★ 11,736 · PYTHON Kornia, a PyTorch library of geometric computer vision routines, added 88 stars to pass 11,700. It packages image processing and geometry operators for what the project calls spatial AI, covering uses from robotics to classical vision tasks. The library dates to 2018 and is still actively maintained.
  4. 4 oraios/serena +57 STARS TODAY · ★ 30,988 · PYTHON Serena, an MCP toolkit that gives coding agents semantic code retrieval and editing, added 57 stars to reach about 31,000. It wires language servers into agents such as Claude Code and Codex so they can navigate a codebase by symbol rather than by raw text. Its language support spans dozens of ecosystems.
  5. 5 eriklindernoren/ML-From-Scratch +36 STARS TODAY · ★ 33,241 · PYTHON ML-From-Scratch, a set of bare-bones NumPy implementations of machine learning algorithms, added 36 stars to reach about 33,200. It rebuilds models from linear regression through deep learning and reinforcement learning as readable Python rather than optimized library code. The repository has not been updated since October 2023.

2026-10-31 · SATURDAY · 13:10 PDT

  1. 1 Orkas-AI/Orkas +1,118 STARS TODAY · ★ 6,050 · JAVASCRIPT Orkas is an MIT-licensed desktop app that runs a local-first AI team: a Commander agent turns a goal into to-dos and coordinates 16 built-in specialists alongside Claude Code, Codex, OpenCode and other harnesses. It added 1,118 stars today and ships projects, scheduled automations and connectors for macOS and Windows.
  2. 2 calesthio/OpenMontage +564 STARS TODAY · ★ 73,909 · PYTHON OpenMontage turns a coding assistant into an agentic video production system, shipping 12 production pipelines, more than 100 tools and over 700 agent-skill and production-knowledge files. The AGPL-3.0 Python project added 564 stars today and lists integrations spanning FFmpeg, Remotion, text-to-speech and image generation.
  3. 3 deepseek-ai/dsh-superagent-prototype ★ 1,883 · TYPESCRIPT DeepSeek published dsh-superagent-prototype, an MIT-licensed TypeScript repository it describes as a step toward a more autonomous agent system. Created today, it carries skills, prompts, examples and tests rather than a packaged product, and has already drawn 1,883 stars.
  4. 4 konlu/story-studio ★ 558 · PYTHON story-studio is an MIT-licensed agent skill that covers AI short-video creation end to end, from picking a topic through to a finished Jianying editing draft. It installs as a SKILL.md package with agents, modules, references and scripts, plus installers for shell, PowerShell and Windows batch.
  5. 5 hevnsnt/CVEngine ★ 546 · PYTHON CVEngine is a GPL-3.0 platform for AI-first external attack surface assessment, pairing continuous discovery with evidence-based vulnerability validation. It builds on nmap, nuclei and bbot, ships a web interface and Docker Compose deployments, and exposes findings to developer agents over MCP.

2026-10-30 · FRIDAY · 13:15 PDT

  1. 1 lightpanda-io/browser +559 STARS TODAY · ★ 38,206 · ZIG Lightpanda is a headless browser written from scratch in Zig for AI agents and automation, not a Chromium fork or a WebKit patch. The project claims 16 times lighter memory use and 9 times faster execution than Chromium on a benchmark that fetches 933 real pages from an AWS m5.large instance. It is AGPL-3.0 and speaks the Chrome DevTools Protocol, so Playwright and Puppeteer scripts can drive it.
  2. 2 MemPalace/mempalace +357 STARS TODAY · ★ 60,152 · PYTHON MemPalace is a local-first memory layer for AI assistants that stores conversation history as verbatim text and retrieves it with semantic search, rather than summarizing or paraphrasing it. The MIT-licensed Python project reports 96.6 percent R@5 on LongMemEval with no API calls, ships an MCP server and uses ChromaDB for vector storage.
  3. 3 ethanplusai/harvey +248 STARS TODAY · ★ 3,448 · PYTHON Harvey is an autonomous sales agent that finds prospects, drafts cold emails, sends campaigns, handles replies and books meetings. It runs on an existing Claude Code or Codex subscription instead of API keys, so the model cost is a seat already paid for. The MIT-licensed Python project added 248 stars today.
  4. 4 langchain-ai/deepagents +60 STARS TODAY · ★ 30,794 · PYTHON Deep Agents is LangChain's open-source agent harness: an opinionated agent that runs out of the box, with defaults tuned for long-horizon, multi-step work and any piece open to being extended, overridden or replaced. The MIT-licensed project is built on LangGraph and shipped for both Python and TypeScript.
  5. 5 NVIDIA-NeMo/Speech +14 STARS TODAY · ★ 18,655 · PYTHON NVIDIA-NeMo/Speech is the speech half of NVIDIA's NeMo framework, a scalable toolkit for researchers and developers working on large language, multimodal and speech models. It covers automatic speech recognition, text-to-speech, speaker recognition, diarization and speech translation. The Apache-2.0 Python repo carries 18,655 stars and was pushed again today.

2026-10-29 · THURSDAY · 13:18 PDT

  1. 1 Donchitos/Claude-Code-Game-Studios +538 STARS TODAY · ★ 28,998 · SHELL Donchitos' repo turns Claude Code into a game development studio, bundling 49 agents and 72 workflow skills into a coordination system that mirrors a real studio hierarchy. The tree carries design, production and studio-test directories plus a project manifest, with topics covering Unity, Unreal Engine and Godot. It added 538 stars today.
  2. 2 thedotmack/claude-mem +459 STARS TODAY · ★ 104,586 · TYPESCRIPT claude-mem records what a coding agent does during a session, compresses it with a model and injects the relevant pieces back into later sessions. The TypeScript project ships plugin directories for Claude, Cursor, Grok and other harnesses alongside an evals suite, and added 459 stars today to pass 104,000.
  3. 3 CopilotKit/OpenBot +240 STARS TODAY · ★ 8,013 · TYPESCRIPT CopilotKit's OpenBot gives each AI coworker a computer of its own - a browser, files and tools - and decides every action before it runs, then records it afterwards. Any AG-UI agent can be plugged in, and the tree holds adapters for LangGraph, CrewAI, Pydantic AI, Mastra and the Claude Agent SDK. It added 240 stars today.
  4. 4 anonnerdfrenzy/survival-llm ★ 746 · PYTHON survival-llm pairs a local AI assistant with an offline survival reference library, running entirely on the user's own machine with no internet connection and showing the source behind every answer. The Python repo includes the library itself, an evals directory and packaging scripts, and was created on October 24.
  5. 5 qybaihe/bla ★ 625 · TYPESCRIPT BLA, short for Blind LLM Agents, is a defensive layer built to stop AI reverse-engineering agents from reading binaries, Electron apps, JavaScript bundles, Python and WebAssembly. Its stated sequence is to poison, exhaust and then mislead the analysing model. The repo was created on October 25.

2026-10-28 · WEDNESDAY · 13:18 PDT

  1. 1 beenuar/AiSOC +453 STARS TODAY · ★ 4,830 · PYTHON beenuar/AiSOC added 453 stars today to reach 4,830. The MIT-licensed project is a self-hostable AI security operations center that ingests security telemetry, correlates threats, and investigates them with LLM agents whose every step lands in a replayable decision ledger. Responses are proposed for a human to approve, and it ships an MCP server for Claude, Cursor and Continue.
  2. 2 expo/skills +212 STARS TODAY · ★ 3,506 · SHELL expo/skills added 212 stars today to reach 3,506. Expo's own MIT-licensed repository packages agent skills for working with Expo projects and Expo Application Services, shipped as a Claude plugin alongside AGENTS.md and CLAUDE.md entry points. It also carries an eval harness and tests for the skills themselves.
  3. 3 macro-inc/macro +106 STARS TODAY · ★ 5,337 · RUST macro-inc/macro added 106 stars today to reach 5,337. The AGPL-licensed workspace combines email, chat, docs, tasks, agents, calls and CRM in one application, with items @-linked across those tools and a shared AI memory behind them. The repository holds the Rust and TypeScript core along with desktop and mobile clients.
  4. 4 anhao/ai-pose-gallery-zh ★ 1,794 · HTML anhao/ai-pose-gallery-zh reached 1,794 stars. The repository is a Chinese localization of a Japanese AI pose reference gallery by @giaaibusi, mirroring all 1,439 records as thumbnails plus full-size images, 2,878 files in total, and translating every metadata field. A zero-dependency static front end adds search, category and viewing-angle filters, grouped browsing and a lightbox.
  5. 5 CopilotKit/OpenDuet ★ 1,232 · TYPESCRIPT CopilotKit/OpenDuet reached 1,232 stars. The MIT-licensed template, built on CopilotKit's OpenBot, is a self-hosted proactive agent that gets a computer of its own and reports back over the web, mobile, Slack or Microsoft Teams. The project labels itself early development and needs a free CopilotKit Intelligence project for threads and memory.

2026-10-27 · TUESDAY · 13:18 PDT

  1. 1 NVIDIA/auto-ontology +75 STARS TODAY · ★ 309 · PYTHON NVIDIA/auto-ontology added 75 stars today to reach 309. The Apache-2.0 Python framework builds semantic, graph-based reasoning layers over existing data, aiming at deterministic query generation, better context retrieval and more explainable output for LLM and agent workflows.
  2. 2 HKUDS/AutoSwarm +41 STARS TODAY · ★ 2,029 · PYTHON HKUDS/AutoSwarm added 41 stars today to reach 2,029. The MIT-licensed Python project runs several coding agents - Codex, Claude Code, DeepSeek Harness, Muse, Grok, WorkBuddy and nanobot among them - as one coordinated swarm, and is documented in an accompanying arXiv paper.
  3. 3 t3dotgg/tstack ★ 1,945 · PYTHON t3dotgg/tstack has reached 1,945 stars since it was created on October 24. The MIT-licensed repository collects nine agent skills for Claude Code and Codex covering the pull request lifecycle: filing a PR, review by three models from different providers, watching CI, rebasing stale branches and merging.
  4. 4 morganlinton/BeautifulRust ★ 1,299 · PYTHON morganlinton/BeautifulRust has reached 1,299 stars since it was created on October 23. The MIT-licensed project ships agent skills that teach coding agents to write, review and port idiomatic Rust, plus a tested clippy configuration, and works with Claude Code, Codex, Copilot, Cursor and Gemini CLI.
  5. 5 andyhuo520/digital-human-koubo ★ 1,049 · TYPESCRIPT andyhuo520/digital-human-koubo has reached 1,049 stars since it was created on October 26. The MIT-licensed TypeScript skill turns a single photo into digital-human talking-head video through Renoise MCP, with five editing styles and 48 packaging components, for Claude Code, Codex, OpenClaw and WorkBuddy.