Apache Maka (Incubating) is a local-first agent workspace developed under the Apache Software Foundation. It inspects projects and runs tools through a shared Runtime Host within a sandbox boundary; tools that leave the sandbox require approval. Model messages, tool calls, tool results, and how a turn ended are stored as recoverable execution facts on the user's machine, while older tool output can be omitted from later prompts without deleting the saved history. Model connections can use cloud APIs, local models, or compatible gateways, with sessions, settings, and run records kept local by default. Maka provides an Electron and React desktop application with streaming sessions, tool timelines, branching, search, recovery, artifacts, and model and sandbox settings. Its TUI/CLI supports work in a project directory and non-interactive turns, while its evaluation surface runs declarative multi-arm experiments across Maka and external subjects. Built-in tools include Read, Write, Edit, Bash, Glob, and Grep; Computer Use and catalog skills are optional. The project is under active development and Apache incubation; the README describes the macOS Apple Silicon desktop build as an early public release and notes that data formats, CLI commands, and experimental capabilities may change.
Archify is an open-source agent skill and Node.js rendering and validation system that turns a codebase or system description into searchable, shareable interactive diagrams. It supports architecture, workflow, sequence, data-flow, and lifecycle diagrams and can be used with Cursor, Claude Code, Codex CLI, and OpenCode. Agents produce typed JSON intermediate representations, which Archify deterministically validates and compiles into self-contained HTML and SVG artifacts. The generated viewer supports node search, optional revision-verified source inspection, upstream and downstream reach tracing, route inspection, semantic-role comparisons, guided stories, themes, presets, and finite motion. Archify can compare two validated snapshots as Before, Delta, and After views, identifying added, removed, changed, moved, and rerouted facts. Its exports include PNG, SVG, WebM, and 1200×630 share cards; it can be installed as a global skill with `npx skills add tt-a1i/archify -g`.
Awesome Agent Skills is an open-source curated collection of more than 1,000 agent skills from official development teams and the community. Each listing provides a description and source link; the collection includes skills from projects such as Anthropic, Google Labs, Vercel, Stripe, Cloudflare, and others. The skills are documented for use with multiple coding-agent environments, including Claude Code, Codex, Gemini CLI, Cursor, GitHub Copilot, OpenCode, and Windsurf.
BetterVoice is an experimental, open-source macOS menu-bar app for local voice dictation with screen context. It transcribes speech locally and captures the full display whenever the user circles an area with the pointer, marking each capture with a blue pointer trail and pulse while preserving multiple references in order. The app inserts the transcript into the selected text field when macOS permits it and pastes captured images when both speech and screen context are available. It supports quick notes and long explanations, configurable microphones, dictation languages, circle-detection thresholds, and keyboard shortcuts. English uses a local Parakeet model by default, while other languages use a separately downloaded multilingual model; an optional English-only grammar-cleanup feature runs locally through quantized ONNX weights.
BookOrbit is a self-hosted library and reading platform for ebooks, PDFs, audiobooks, and comics. It provides web readers and synchronizes reading progress, highlights, annotations, and reading status between BookOrbit, Kobo devices, and KOReader. The platform includes library management with metadata enrichment, multiple libraries, collections and rule-based saved filters, multi-user accounts with OIDC/SSO, OPDS access, Send-to-Kindle delivery, and browser uploads. It also provides reading statistics, goals, achievements, searchable annotation exports, and synchronization with services including Hardcover, Readwise, and StoryGraph. The project is designed to run on infrastructure controlled by the operator.
book-to-skill is an open-source agent skill that converts technical books, document folders, or collections of source files into a unified skill for GitHub Copilot CLI, Amp, or Claude Code. It accepts PDFs, EPUBs, office documents, plain text, folders, globs, and file lists; the videos also describe OCR support for scanned PDFs. The tool distills source material into structured knowledge rather than a single summary, extracting frameworks, decision rules, anti-patterns, and per-chapter Markdown files. It also generates a core SKILL.md with a chapter index, plus glossary, patterns, and cheatsheet files. Agents load the relevant chapter and supporting files on demand, allowing answers to be grounded in the source content without placing the entire book in the context.
CarWatch is an open-source, offline in-car AI agent from ThinkOffApp that runs on a Raspberry Pi 5 with a locally hosted language model. It joins chat rooms as a vehicle agent, answers questions from the car's owner's manual using lexical retrieval-augmented generation and page citations, and reads live vehicle and system data while stating when information cannot be sensed. A continuous energy-based voice listener and whisper.cpp provide local speech input without a wake word, with responses played through the car's speakers. The system can monitor OBD and engine data, send departure and arrival messages, trip summaries, and dashcam clips, and expose a phone dashboard for approvals, replies, model selection, and maintenance. The repository describes systemd-managed services for the model server, room agent, voice listener, dashboard, and engine watcher, with Python standard-library code and no cloud subscription required.
Coolify is an open-source, self-hostable platform as a service for managing servers, applications, and databases on users' own hardware. It deploys static sites, full-stack applications, databases, and one-click services on VPSs, bare-metal machines, Raspberry Pis, and other servers, using an SSH connection rather than requiring a vendor-managed cloud environment. It uses Docker to build and deploy applications and supports Git-based deployments, preview deployments, SSL, backups, monitoring, multi-server setups, Cloudflare Tunnels, and an API. The project's configurations remain on the user's servers, so running resources can continue to be managed without Coolify, although its automations are lost. Coolify can be installed with a shell script and is also offered as a paid cloud version for users who do not want to self-host the Coolify server. The repository describes it as an alternative to Heroku, Netlify, and Vercel.
GoogleTest is Google's open-source C++ testing and mocking framework, maintained in the repository that merged the formerly separate GoogleTest and GoogleMock projects. Based on the xUnit architecture, it automatically discovers and runs tests and provides built-in and user-defined assertions, including equality, inequality, exception, fatal and non-fatal failure, and death-test assertions. It supports value-parameterized and type-parameterized tests, as well as options for running individual tests, controlling test order, and executing tests in parallel. The 1.18.x branch requires at least C++17; documentation is provided through the project's GitHub Pages site.
Impeccable is an open-source design skill for AI coding agents, developed by pbakaus. It provides a shared design vocabulary with 23 commands, including init, shape, critique, audit, polish, harden, animate, typeset, layout, and live. The init command gathers product and design context into PRODUCT.md and DESIGN.md files, while other commands review, plan, modify, or refine a frontend project. Its CLI and browser extension run 61 deterministic detector rules for common AI-generated frontend design problems without an LLM or API key; it also supports LLM-based critique checks. The live command provides browser-based visual iteration, and the package can be installed from a project root with npx impeccable install.
Marin is an open-source research program, software platform, and community for developing foundation models, with a focus on large language models. It covers data curation, transformation and filtering, tokenization, pretraining, posttraining, and evaluation, while recording processes, experiments, decisions, artifacts, and failed experiments from raw data through final model. Marin experiments are defined as dependent steps executed in topological order, similar to a Makefile, and the platform can scale from small language-model runs to large TPU deployments. The project has also been used for audio-text, DNA, and protein models and provides documentation, experiment code, checkpoints, and reusable training pipelines.
OpenAI Codex is an AI coding agent from OpenAI available as a command-line tool (Codex CLI) that helps developers produce software. It can be used alongside Gemini for adversarial audits of software requirements and implementation plans, listed as a supported coding-agent or model option in several projects, and its logs can be joined with task and test evidence. The Codex CLI can also receive and answer requests from the Penako canvas.
OpenHuman is a local-first personal AI system from tinyhumansai that combines persistent memory, agent orchestration, and research tools. It stores a user's data as scored Markdown trees in SQLite on the local machine and mirrors the result to an editable Obsidian vault; its TokenJuice component compresses tool output before it reaches the language model. Its orchestration layer runs checkpointed, durable agent workflows and worker fleets on graphs, with triggers, approvals, steering, halting, and replay support. The system also provides web search, scraping, coding tools, a browser, native voice through in-process Whisper, model routing, messaging integrations, and support for provider keys or fully local Ollama models. The repository describes it as an early beta under active development.
OpenViking is an open-source context database for AI agents developed by Volcengine. It unifies agent memories, resources for knowledge retrieval, and skills in a virtual filesystem accessed through the viking:// protocol, allowing agents to browse context with filesystem-style operations such as ls, tree, and find instead of querying an opaque vector store. Content is processed into three loading tiers: L0 abstracts for relevance checks, L1 overviews for planning, and L2 full details loaded on demand. Recursive retrieval first locates a relevant directory through vector search and then drills down through its hierarchy, preserving the surrounding context and recording the browsing trajectory for inspection and debugging. After a session is committed, OpenViking asynchronously extracts user preferences and agent experience into long-term memory. The project also provides an OpenViking Studio browser playground, documentation, and a live demo.
ThreeUI is a free, open-source, login-free catalog and React component library for interactive Three.js, WebGL, shader, and UI components. Its Community edition provides browsing, search, themes, responsive layouts, live renderers, controls, variant pickers, source-code tabs, and complete Community component source code, while excluding Pro and Beta components and authentication, checkout, and Pro implementation source. The public library can be installed from npm as @designcodeio/threeui and imported as React components with shared styles through the package root or component subpaths, or run locally with Vite from the repository. Components that render full HTML documents require runtime assets to be copied into the host application's public directory. Application code and Community component code are MIT licensed; bundled fonts use the SIL Open Font License 1.1 and bundled Three.js runtime files use the MIT license.
vgpu is an open-source TypeScript library for WebGPU, developed by Vercel Labs, covering typed shaders, 3D scenes, GPU tensors, neural networks, and mathematical visualization. WGSL files can import and export declarations like TypeScript modules; reflection tracks binding names, types, and layouts so shader bindings do not require handwritten declarations. Its tiny GPU-first API uses a single context returned by init(), which is passed explicitly to rendering entry points. The same API runs in browsers, headless Node through a Dawn-backed device, and a deterministic software mock for tests and continuous integration. Explicit frame passes handle targets, effects, clears, and draws without hidden scene-graph state. The package includes a CLI for documentation, examples, and shader validation, and the video describes an MCP endpoint for agent access.
Vorssaint is a free, open-source Swift/SwiftUI macOS menu bar toolkit that combines utilities including per-application volume and audio controls, output and microphone switching, system monitoring, network statistics, app and window switching, window snapping, Dock previews, clipboard history, text snippets, file staging, screen capture, display and energy controls, media tools, cleaning, and app removal. It is modular: users can install or remove complete features, with removed features stopping loading while retaining their settings for later reinstallation. The toolkit also provides configurable shortcuts, panels, a command bar, and optional bundles for common setups. Vorssaint runs locally without an account, analytics, tracking, subscription, or required network service; its volume mixer requires no audio driver or setup, and selected features request only the relevant macOS permissions, including accessibility, screen recording, audio recording, microphone, camera, calendars, notifications, and disk access.
walgit is a Rust-based Git server distributed as a single binary that stores repositories in S3-compatible object storage or Google Cloud Storage, without a database, elected leader, or important local state. The object store acts as the source of truth: pushes are written as immutable objects and made visible by atomically rewriting a manifest with compare-and-swap, while local repositories serve as disposable caches. Replicas can point to the same bucket and serve repositories without coordination; compaction publishes compacted packs back into the log. It provides Git smart HTTP fetch and push, bundle-URI clones served as static files, Git LFS, a repository-browsing web UI, a JSON API with an SDK, per-repository push policies, webhooks, and token or OIDC authentication.
Workout Guide is an open exercise illustration library and framework-neutral npm package by Bryl Lim. It provides structured metadata for 302 exercises, three consistent frames per exercise, transparent normalized assets, and typed APIs such as `getExercise`, `searchExercises`, and `getAssetUrl`; its searchable static gallery includes exercise detail pages and integration guidance. The repository is an npm-workspace monorepo containing the package API, canonical manifest, asset files, gallery site, documentation, and deterministic catalog import and validation scripts. Code and documentation are licensed under MIT, while the visual assets are licensed under CC BY-SA 4.0, including artwork derived from Everkinetic.
An open-source agent skill that turns a codebase or system description into searchable, shareable, self-contained HTML diagrams.
An open-source agent skill that converts technical books, folders, or documents into on-demand skills containing frameworks, decision rules, anti-patterns, and per-chapter files.
An open-source TypeScript library covering WebGPU shaders, 3D scenes, GPU tensors, neural networks, and mathematical visualization with typed shader imports and cross-runtime execution.
Searchable transcript of Trending Open-Source GitHub Projects : Archify, Marin, Codex, Coolify & CarWatch #287 — ManuAGI - AutoGPT Tutorials (15:50). Search for a phrase, then click its timestamp to jump straight to that moment in the video.
Captions sourced from the original video on YouTube, published by ManuAGI - AutoGPT Tutorials. The video, its captions and all related intellectual property remain the property of their respective owners; AINotes claims no ownership. Provided for research, accessibility and search — see the Transcript Notice and Copyright Policy.
00:00 Every week, developers release powerful open-source GitHub projects that make coding, research, and daily work faster. This is your weekly GitHub project update video, covering top trending open-source GitHub projects this week. From AI agent skills to self-hosted platforms and local first tools, you'll discover useful and trending developer tools worth adding to your stack.
00:21 Without wasting time, let's get started. >> Before we jump into today's project updates, here's a quick announcement for everyone. We've launched a brand new YouTube channel called AI Agent Studio dedicated entirely to AI Agent projects, tutorials, and tools. So, if you're interested in staying up to date with the latest AI Agent open source projects, learning how to build your own agents, or exploring cuttingedge agent frameworks, make sure to check it out.
00:50 Subscribe now to get weekly videos, in-depth guides, and real time project breakdowns. The link is right there in the description. Don't miss it. All right, let's get into today's video. >> Project number one, Archify. Turn a codebase into verifiable interactive diagrams. Archify is an open-source agent skill that turns a codebase or system description into an interactive diagrams.
01:13 It works in cursor, clawed code, codeex cli, and open code. You describe a system or a repository, and it produces a self-contained HTML map you can search and share. It covers five diagram types, architecture, workflow, sequence, data flow, and life cycle, and keeps nodes grounded in authored facts rather than inventing topology. It validates diagrams before exporting PNG, SVG, and webm.
01:39 Map your system. Project number two, fonts, scientific agent skills, 563 science skills that turn agents into scientists. Scientific Agent Skills is an open-source library of 163 science skills that turn any AI agent into a research assistant. It spans biology, chemistry, medicine, and drug discovery, wrapping rdkit, scanpie, and pytorch lightning, and unifying access to 100 databases, pub, chemil, uniro, clinical trials.gov, and more.
02:11 Each skill ships documentation, examples, and tests. And every skill is security scanned. It works with cursor, clawed code, codecs, and any agent skills host. Installable in one command. Turn your agent into a scientists. Project number three, Marin. Open-source framework for reproducible foundation model research. Marin is an open-source framework for researching and developing foundation models from data to a trained model.
02:37 Its key feature is reproducibility. Every step, data curation, filtering, tokenization, training, and evaluation is recorded, including failed experiments. So the research process stays transparent. You define experiments as dependent steps run in tupilological order, like a make file, scaling from a tiny CPU model to large TPU pods. Built in Python, it trains models like llama, Quinn, and DeepSeek.
03:05 Install it and train your own. Project number four, Google Test. Google's C++ testing and mocking framework. Google Test is Google's open-source C++ testing and mocking framework. Merging Google Test and Google Mock projects based on the XUnit architecture. It discovers and runs your tests automatically with no manual registration and offers many assertions plus your own custom ones.
03:26 It supports death tests, fatal and non-fatal failures and value and type parameterized tests and can run tests individually, in order, or in parallel. It builds with CMake or Basil and powers Chromium, LLVM and Open CV. Add it and start testing. Project number five, Miss Codeex. Lightweight coding agent that runs in your terminal. Codex CLI is a lightweight open-source coding agent from OpenAI that runs locally in your terminal.
03:56 You give it a task and it reads, edits, and runs code. You sign in with your chat GPT plan plus pro business or enterprise or use an API key. It's available in your IDE, a desktop app, and the cloud as codeex web. Built in Rust, it installs with one command, Homebrew or npm across Mac OS, Linux, and Windows. Install it and code from your terminal. Project number six, dashaka, local first open-source AI agent workspace.
04:25 Macaka is a local first open-source AI agent workspace. Beyond answering questions, it inspects projects, runs tools like read, write, and bash under a permission policy. Sessions, settings, and run records stay on your machine, and you choose the model, cloud API, local model, or gateway. Every message and tool call enters a runtime event log that drives sessions, context, and recovery.
04:49 Reach it from a desktop app, TUI, or CLI. It's an early release. Build it and put it to work. Project number seven, the open human local first personal AI memory orchestration and research. Open human is a local first open-source AI that combines memory orchestration and research. It builds a private memory as scored markdown trees in local SQLite mirrored to an Obsidian vault and fed by 100 plus integrations.
05:17 It orchestrates fleets of agents on durable graphs connecting claude codecs and others over encrypted agentto agent sessions. It adds web search, browser, and 17 messaging channels, and a privacy mode keeps inference on your machine. It's an early beta. Install it and build your brain. Project number eight, Awesome Agent Skills. Curated list of 1,000 plus agent skills.
05:41 Awesome Agent Skills is an open-source curated list of over 1,000 agent skills from official dev teams and the community. Rather than bulk AI generated entries, it collects real skills published by teams like Anthropic, Google, Verscell, Stripe, Cloudflare, Sentry, HuggingFace, Figma, OpenAI, and Microsoft alongside community contributions. Each entry links to the skill with a description and its source, and everything works with Claude Code, Codeex, Gemini CLI, Cursor, GitHub Copilot, and more.
06:13 Browse it and equip your agent. Project number nine, Book Orbit. Self-hosted library for ebooks, audiobooks, and comics. Book Orbit is a self-hosted open-source library and reading platform for ebooks, PDFs, audiobooks, and comics. Its built-in web readers open EPUB, MOI, PDF, CBZ, and audiobook formats with no plugins, and progress and annotations sync birectionally across Coobo, KO Reader, and the web.
06:41 Highlights merge into one searchable hub you can export and it enriches books with metadata from Google books, Goodreads and more. It supports multiple users with OIDC, OPDS and send to Kindle. Deploy it with Docker and own your library. Project number 10, Open Viking file system style context database for AI agents. Open Viking is an open-source context database for AI agents that unifies memory, knowledge rag, and skills.
07:10 Instead of a vector store, it organizes everything under a file stem, mapping memories, resources, and skills to Viking D-sho paths and agent browses with file stem commands. Context is tiered into abstract, overview, and detail layers loaded on demand. Retrieval recurses through directories, and every retrieval trajectory is visible. Sessions extract memory automatically from Vulks engine.
07:32 It installs with pip. Install it and give your agent context. Project number 11, book tokill. Turn any book into an ondemand agent skill. Bookto-skill is an open-source agent skill that turns any technical book, folder, or documents into a skill your agent loads on demand. You point it at a PDF, EPUB, docx, or folder, and it distills it into frameworks, decision rules, antiatterns, and per chapter files.
08:00 Structure, not a summary. Your agent reads the relevant chapter and answers from real content so it doesn't hallucinate. It runs locally and works in Claude Code, Copilot CLI, and AMP. Install it and study any book. Project number 12, VGPU cross Runtime WebGPU library with type shader imports. VGPU is an open-source TypeScript library for web GPU covering shaders, 3D scenes, GPU tensors, neural networks, and math visualization.
08:30 Its wgsl files import and export like typescript modules and reflection keeps bindings and types correct with no codegen. One init call returns a GPU context that every entry point takes. The same code runs in the browser headless node through dawn and a deterministic mock for tests. A CLI and MCP endpoint make it agent ready. Install it and start rendering.
08:55 Project number 13. Koolifi self-hostable paz alternative to Versell and Heroku. Koolifi is an open-source self-hostable pad and an alternative to Versell, Heroku, and Netlefi. It manages your own servers, VPS, bare metal, or Raspberry Pies over nothing but an SSH connection, letting you deploy static sites, databases, full stack apps, and 280 plus one-click services.
09:20 It handles Let's Encrypt SSL, Git deploys from GitHub, GitLab, and Gideia, S3 backups, and offers a full API. There's no vendor lockin since every config lives on your server. Install it and deploy on your own servers. Project number 14, Impeccable, design skill that sharpens AI coding agents. Impeccable is an open-source design skill that makes AI coding agents better at front-end design.
09:48 It installs into clawed code, cursor, codeex, gemini, and more. Adding 23 commands, audit to find issues, polish for cleanup, distill to remove complexity, critique for a full review, each able to focus on a specific area. It runs 59 deterministic checks that catch the usual AI design tells like inter everywhere and purple to blue gradients and iterates live in the browser.
10:13 Install it and sharpen your interfaces. Project number 15, Vorsant. Free open-source Mac OS menu bar toolkit. Vorsant is a free open-source Mac OS menu bar toolkit that bundles utilities Mac users usually buy one by one per app volume, a system monitor, an app switcher, Windows snapping, clipboard history, text snippets, a file shelf, and an uninstaller behind a single icon with no account, telemetry, or subscription.
10:41 It's modular. Uninstall a feature and it stops loading and using energy and reinstalling restores your settings. A command bar runs any action from one shortcut. Download it and reclaim your menu bar. Project number 16. 3UI community, a free browsable catalog of live WebGL shader components. 3UI community is the open-source login free edition of 3UI, a React and 3JS component catalog.
11:08 It solves the problem of finding ready-made shader-driven UI pieces by giving developers a browsable grid of components with live renderers, controls, variant pickers, and source code tabs, all running in the browser through WebGL. It shares its application shell, navigation, search, and themes with the main 3 UI project, but strips out pro and beta components while keeping every free variant fully working.
11:32 Maintainers sync this community subset from the main project through a script that filters restricted content and rebuilds the public source bundle. It runs locally with npm and vit. This suits front-end developers who want interactive shader components without an account. Clone the repository and run it locally to explore the catalog. Project number 17, Walgit, a git server that is one binary over object storage.
11:55 Walgit is a rustbased git server that hosts repositories directly on an S3 or GCS bucket with no database and no local state that matters. It solves the pain of scaling git hosting by treating the bucket as the source of truth and every server machine as a disposable cache. A push writes an immutable object and flips a small manifest through a compare and swap.
12:19 So any instance can accept rights safely without a leader or quorum. It serves smart HTTP git bundle URI clones, Git LFS, a web UI, a JSON API with an SDK per repository push policy and web hooks, and supports token or OIDC authentication. This suits teams hosting large or many repositories who want Git infrastructure that scales without fleet management.
12:44 Point it at a bucket and push your first repository to see it work. Project number 18, Workout Guide. An open exercise illustration library with a typed npm package. Workout guide is an open-source exercise illustration library packaged as a framework neutral npm module, giving developers structured access to 302 exercises with three consistent animation frames each.
13:08 It solves the problem of sourcing consistent licensed exercise artwork by bundling normalized PNG assets with typed functions like get exercise, search exercises, and get asset URL. The package works with React Native through direct asset imports and require calls alongside a searchable static gallery site built with Astro. The repository is an npm workspace mono repo with scripts to import and validate catalog data from source exports.
13:35 This suits developers building fitness apps who need ready-made exercise visuals and metadata instead of sourcing their own. Install the package and pull an exercise into your app to see it work. Project number 19, Carwatch. Your car joins the chat as an offline AI agent. Carwatch turns a Raspberry Pi 5 into an AI agent that rides in your car and joins chat rooms on group mind as a named participant working fully offline.
14:01 It solves the problem of getting car status and safety relevant answers without relying on the cloud. Running a local language model on the Pi for grounded ondevice reasoning. It answers questions from the car's own owner's manual using ondevice retrieval with page citations, reads live machine state like temperature and network status, and listens continuously for hands-free voice input through ani speech pipeline.
14:28 It reads engine data over the OBD port, pulls dash cam clips, and syncs with a phone dashboard the car itself serves. It self-updates from this repository, and reaches you through a dialout tunnel, even without direct network access. This suits hobbyists and developers building offline edge AI hardware projects for their vehicle. Clone the repo and set it up on a Pi to bring your car into the chat.
14:50 Project number 20, Better Voice. Voice dictation that captures the screen you point at. BetterVoice is an open-source Mac OS menu bar app that transcribes speech locally while letting you circle anything on screen to attach as visual context. It solves the problem of voice dictation losing the context behind what someone is describing by pairing a spoken note with screenshots of exactly what the pointer circled.
15:13 Holding a key starts a short recording and a longer shortcut supports extended dictation with transcripts inserted directly into the selected text field when Mac OS allows it. Speech to text runs locally through fluid audio, so audio never leaves the machine and captured images and text also copy to the clipboard for pasting elsewhere. It targets Apple Silicon on Mac OS 14 and later.
15:35 This suits developers and writers who want dictation tied to on-screen references instead of plain text. Download the app and circle something on your screen to try it. Thanks for watching. See you in the next update.