Trending Open-Source GitHub Projects : CloddsBot, Helium, Archify, OpenResearch & MacTap #292
Read the Portuguese translation →
Author: ManuAGI - AutoGPT Tutorials · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: Trending Open-Source GitHub Projects : CloddsBot, Helium, Archify, OpenResearch & MacTap #292 · Channel: ManuAGI - AutoGPT Tutorials · 15:42 · Watch on YouTube
Topics: AI agents, Task and workflow automation, Open source and software ecosystems, Privacy and data protection
Summary
Section titled “Summary”This weekly roundup introduces 20 open-source projects for AI agents, developer workflows, privacy, automation, and creative tools. Highlights include Claudesbot, an autonomous trading agent; OpenResearch, which coordinates parallel research agents; Archify, which creates interactive diagrams from codebases; and Helium, a privacy-focused Chromium browser. The video also covers local-first AI workspaces and context tools, Android UI test automation, and MacTap, which turns knocks on a MacBook into shortcuts. It closes with a range of tools for testing, music, motion graphics, and generative drawing.
Key points
Section titled “Key points”- The video presents 20 open-source projects spanning AI agents, developer tools, desktop utilities, and creative applications.
- Claudesbot is described as a self-hosted AI trading agent that operates across multiple markets and can be controlled through messaging platforms or web chat.
- OpenResearch runs parallel research agents through an iterative loop of proposing ideas, editing code, running experiments, and reviewing results.
- Archify turns codebases or system descriptions into interactive diagrams and supports several diagram types and export formats.
- Helium is a Chromium-based browser focused on privacy, with tracking components removed and ad blocking included.
- Several projects emphasize local-first or self-hosted use, including OpenResearch, OpenViking, Macaka, and Everroom.
- The roundup includes automation tools such as natural-language Android UI testing and a JavaScript browser environment for tests.
- MacTap maps physical knocks on a MacBook to shortcuts and actions, running on device according to the video.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | Introduction and channel announcement | The presenter introduces the weekly open-source project roundup and announces the AI Agent Studio channel. |
| 01:02 | Projects 1–5: sideloading, chat CRM, trading, networking, and research | The video covers an iOS sideloading tool, a self-hosted chat CRM, Claudesbot, a modular TCP tunnel, and OpenResearch. |
| 04:16 | Projects 6–10: diagrams, writing, agent context, Mac tools, and LLM routing | The presenter introduces Archify, Humanizer, OpenViking, a macOS menu bar toolkit, and the Switchyard proxy. |
| 07:07 | Projects 11–14: local AI workspace, Android testing, JavaScript testing, and Helium | This section presents Macaka, Google’s natural-language Android UI testing tool, Hagdum, and the Helium browser. |
| 10:19 | Projects 15–17: AI brand auditing, traceable context, and music | The roundup covers Newbio for auditing AI brand visibility, Everroom for traceable agent context, and the Spun Linux music player. |
| 13:04 | Projects 18–20: explainer videos, generative drawing, and MacTap | The final projects create coded explainer films, provide P5.js drawing brushes, and map MacBook knocks to shortcuts. |
Quotes
Section titled “Quotes”Every week, developers release powerful open-source GitHub projects that make coding, automation, and daily workflows faster. — 00:00
It automates the research loop. — 03:43
keeps nodes grounded in authored facts rather than inventing topology. — 04:16
A privacy first Chromium browser with built-in ad blocking. — 09:17
MacTap turns knocks on your MacBook into shortcuts. — 14:37
Concepts and entities
Section titled “Concepts and entities”CloddsBot, OpenResearch, Archify, OpenViking, Helium, MacTap, AI Agent Studio, WAHA
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 Every week, developers release powerful open-source GitHub projects that make coding, automation, and daily workflows 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. Without wasting time, let’s get started. Before we jump into today’s project updates, here’s a quick announcement for everyone, we’ve
00:30 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. 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.
01:02 >> Project number one, Lie Loader. User-friendly sideloadader for iOS apps. Live loader is an open-source user-friendly sideloadader for installing apps on your own iOS device with your Apple ID. Instead of the usual manual steps, it installs side store or live container with side store, signs it with your certificate, and places the pairing files for you automatically. It’s a signed notorized desktop app for Mac OS, Windows, and Linux built on a Rust sideloading library with log viewing and many language localization.
01:34 Newer builds add wireless install to Apple Vision Pro. Download it and sideloadad side store. Project number two, DeskCom CRM. Self-hosted AI sales CRM for selling by chat. Deskcom CRM is an open-source self-hosted CRM for businesses that sell by chat, an alternative to Como, Octtoesk, and Intercom. It combines a sales pipeline with AI agents and WhatsApp through WAHA. So, leads flow from chat and forms. The AI replies and books appointments, and a human can take over to stop the bot. It’s MCP ready and
02:08 multi-tenant with automations, scheduling, Google calendar sync, and LGPD compliance. One command installs the stack on a VPS. Deploy it and sell by chat. Project number three, Claudesbot, self-hosted autonomous AI trading agent across many markets. Claudesbot is an open-source self-hosted AI trading agent built on Claude that operates across 1,000 plus markets. Poly Market, Kalshi, Binance, Hyperliquid, Salanaex, and several EVM chains. It
02:38 pulls real-time and historical market data, scans for opportunities, places limit and market orders with smart routing, and tracks P&L and risk. You run it yourself, control it from 21 messaging platforms or a built-in web chat, and expose it to Claude through an MCP server. Install it and run your own agent. Project number four, Banch Open Flux. TCP tunnel with pluggable transports for research. Open Flux is a network stack research tool built around a TCP tunnel with pluggable transports. It carries a TCP connection over a
03:12 transport you choose, and its transport layer is modular, so you can swap or add the way traffic is moved without changing the rest of the stack. Useful for experimenting with and comparing network behaviors. It’s written in Go with companion rappers for Android and desktop and its documentation is in English and Russian. Clone it and run a tunnel. Project number five, Elm research. Run parallel autonomous research agents with any model. Open research from Alphacy. Runs parallel research agents with any model. It
03:43 automates the research loop. An agent proposes an idea, edits the code, launches an experiment, inspects the results, and decides what to try next. While multiple agents explore directions and an experiment tree keeps their lineage, the same committed snapshot runs locally over SSH or on Slurm, Kubernetes, Ray, Modal, and more. It’s local first, storing projects in SQLite without publishing your code. Install the CLI and start researching. Project number six, DAC, Archify. Turn a codebase into verifiable interactive
04:16 diagrams. Archify is an open-source agent skill that turns a codebase or system description into an interactive diagrams. It works in cursor, clawed code, codec 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. [clears throat] Map your
04:51 system. Project number seven, humanizer. Agent skill that removes AI tells from writing. Humanizer is an open-source agent skill that removes signs of AI generated writing from text. It checks your writing against 35 patterns that tells of machine writing and rewrites them into pros that sounds human. A no fabrication rule keeps it from inventing facts. Specifics can only come from your source or you and it flags unsupported claims. You can supply a voice sample to override its defaults and run it on pasted text, a file or inline. Install
05:25 it and delop your writing. Project number eight, 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. Instead of a vector store, it organizes everything under a file stem, mapping memories, resources, and skills to Viking file system ash in agent browsers with file stem commands. Context is tiered into abstract overview and detail layers loaded on demand. Retrieval recurses through directories and every
05:57 retrieval trajectory is visible. Sessions extract memory automatically from Vulc engine. It installs with pip. Install it and give your agent context. Project number nine, 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
06:32 subscription. 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 10, Switchyard. Rust proxy that roots and translates LLM traffic. Switchyard is an open- source Rust proxy and library for LLM traffic from Nvidia Nemo. It translates between OpenAI chat, anthropic messages, and OpenAI responses. So cloud code or codeex keeps its native API while requests are served
07:07 by VLLM, NVIDIA NIM, Alama or any OpenAI compatible endpoint. It roots with classifier stage and escalation strategies or your own and records Prometheus metrics. Run it as a launcher, standalone server or Rust library. It’s pre-alpha. Install it and route your traffic. Project number 11, bench Apache Maka local first open-source AI agent workspace. Macaka is a local first open-source AI agent workspace. Beyond answering questions,
07:38 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. Reach it from a desktop app, TUI, or CLI. It’s an early release. Build it and put it to work. Project number 12. Outer natural language Android UI test automation from Google. Artmees is an open-source tool from
08:13 Google that turns natural language instructions into Android UI automation. You describe a test scenario and it generates a step-by-step plan, drives a device or emulator, navigating and profiling performance, then returns a report with metric tables and data. It locates elements using accessibility trees, OCR, and visual models. So, it handles canvas, compose, and flutter screens. It plugs into AI assistants like anti-gravity, codecs, and clawed code over MCP. Install it and automate your Android tests. Project number 13,
08:47 Pacid Dum. A browser environment for JavaScript without the graphical interface. Hagdum is a JavaScript library that recreates how a web browser works minus the actual visible interface. It solves the problem of testing or running browser dependent code in node without opening a real browser window. It works by implementing DOM features like custom elements, shadowdom, mutation observer, tree walker, and the fetch API directly in JavaScript. It connects with tools like vite test, gest, bun, and testing
09:17 library and supports frameworks such as React, Vue, Angular, and Spelt. This is useful for developers who need fast, lightweight test environments for front-end code. Explore its documentation to set it up as your test environment. Project number 14, Helium. A privacy first Chromium browser with built-in ad blocking. Helium is a Chromiumbased web browser built around privacy and simplicity. It solves the problem of browsers collecting user data and cluttering the experience with bloat. It works by building on ungooled
09:48 chromium, stripping out tracking components and adding unbiased ad blocking through its own fork of Ublock origin. It runs natively on Mac OS, Linux, and Windows with dedicated packaging repos for each platform. Helium also includes separate repos for its web services and onboarding components, all open source. This matters for people who want a fast, familiar browser without hidden data collection. Download Helium from helium.comput and try browsing without the noise. Project number 15. Newbio,
10:19 open-source auditing tool for AI brand visibility. Newbio is a self-hosted tool that checks how AI models talk about a brand. It solves the problem of not knowing whether AI recommends your product or pushes competitors instead. It works by taking a domain, identifying the brand and its competitors, sending confirmed customer questions to provider APIs, then analyzing the answers and sources behind them. It connects with Open Router, OpenAI, Anthropic, Google Gemini, Perplexity, Deepseek, and OpenAI compatible endpoints using your own API
10:52 keys. It runs locally through Docker or NodeJS, and every conclusion links back to the original AI answer. This matters for founders who want a clear evidence-based view of how AI represents their product. Run a domain through Newbie Geo and see what AI actually says about it. Project number 16, Everroom Room, a local first workspace that keeps AI agent context traceable. Everroom is a desktop platform that gives AI agents organized, trustworthy context to work with. It solves the problem of agents losing track of sources, decisions, and
11:25 prior reasoning across a project. It works by pulling in files, repositories, conversations, and connected apps, then normalizing that material into an evidence ledger with clear provenence. Content is organized into rooms, scoped workspaces where agents only see the sources and tools relevant to that project. It runs locally on Mac OS using Electron, a Fastify gateway, and SQLite, keeping data, credentials, and documents on the user’s device. This matters for developers building agent workflows who need every AI generated summary
11:57 traceable back to its source. Explore the room and agent setup to see how it keeps context accountable. Project number 17, Dish Spun, a CD-shaped music player for Linux desktops. Spun is a desktop music player for Linux that solves the problem of switching between apps to play local files and stream Apple Music. It works by turning album artwork into a spinning disc on screen using a quiet interface styled after material design 3. Local music plays directly through QT multimedia while Apple Music streams through Citer’s
12:29 local API which handles authentication and playback. Spun connects to Citer on a local port to pull library search and Q data. It builds from source using CMake, QT6 and Tag Lib on Linux with Cachios and Arch Linux as the tested environments. This suits Linux users who want one player for both local and Apple Music libraries. Build it from source and start spinning your library. Project number 18. Anything to explainer turns any topic into a coded motion graphics video. Anything to explainer is a skill for clawed code and codecs that turns a
13:04 topic or document into a narrated explainer film. It solves the problem of building an animated explainer video without a video editor or design team. It works by having an AI agent research the topic. Write narration. Generate voice over with frame accurate timing. Storyboard each shot. Then dispatch parallel build agents that write the animation in motion using React and TypeScript. Every frame is code, not stock footage or generative video, so any shot can be fixed by editing its file. It renders on CPU through headless
13:34 chromium, no GPU needed, and supports Chinese or English narration with matching typography and TTS voices. This suits developers and creators who want data grounded explainer videos without hand animation. Point it at a topic and let the agent build the film. Project number 19. Cool Brush A. A browser drawing app with six expressive P5.js brushes. Cool Brush A is a browserbased drawing app built with P5.js offering six distinct brushes on one
14:05 shared canvas. It solves the problem of switching between brush styles without reloading the page or losing existing artwork. It works by giving each brush its own JavaScript module settings and preview. All pulled from a shared registry that keeps the canvas and history intact across switches. Brushes range from flowing staff ribbons and wet ink schematics to animated crowds and model kit sprus each with adjustable size, density, and style options. It runs locally in the browser using Vite for development and bundling with p5.js
14:37 installed locally rather than loaded from a CDN. This suits creative coders and illustrators who want playful generative brush effects without external drawing software. Open the app and start switching between brushes to see what each one draws. Project number 20. MacTap turns knocks on your MacBook into shortcuts. MacTap is a Mac OS menu bar app that turns physical taps on a MacBook’s chassis or the desk beneath it into keyboard shortcuts. It solves the problem of reaching for a key combo when a simple knock could do the job. It
15:10 works by using the MacBook’s built-in motion sensor to classify one, two, or three knocks, then mapping each pattern to an action like copy, paste, screenshots, or media control. Left and right edges of the chassis trigger separate sets of actions and presets like coding, capture, and focus adjust the mappingss for different tasks. It runs entirely on device with Apple Silicon Max on Mac OS 14.6 or later, uploading nothing. This suits Mac users who want quick physical shortcuts without memorizing key combinations.
15:42 Install it from the releases page and start knocking. Thanks for watching. See you in the next update.