Audar-ASR-V1 is a family of Arabic-first generative speech-recognition models developed by AudarAI. It treats transcription as audio-conditioned next-token prediction with a language-model decoder, rather than a CTC or transducer objective, using a Whisper-style 128-mel audio encoder and a Qwen3 decoder with a 30-second context. The models cover Modern Standard Arabic, major Arabic dialects, code-switched Arabic-English, English, and 30 languages overall. The Flash tier is intended for real-time, edge, on-device, or offline use, while Turbo targets lower error on difficult dialectal and long-form audio. Both tiers share an architecture and prompt interface and can be run through Transformers, GGUF with llama.cpp, or vLLM; the repository provides model pointers, benchmarks, and inference examples. The published models use the AudarAI Open v1.0 and AudarAI Community v1.0 licenses.
Code Review Graph is an open-source, local-first code intelligence tool that provides a command-line interface and MCP server. It builds a persistent structural map of a codebase with Tree-sitter, representing relationships such as functions, classes, imports, calls, inheritance, and test coverage, while tracking changes incrementally. The graph supplies focused repository context to AI coding tools so they can read relevant code instead of repeatedly scanning large portions of a project. It also supports code-review workflows such as blast-radius analysis and risk-scored pull-request reviews. The Python package can be installed with pip or pipx; its install command detects supported AI coding platforms and configures MCP settings, hooks, skills, and graph-aware instructions.
DeepTutor is an open-source, self-hostable AI tutoring platform developed by HKUDS for lifelong personalized tutoring. It provides a web application and command-line interface for interactive learning, problem solving, quiz generation, deep research, visualization, and mastery practice, with persistent memory and learning state shared across knowledge bases, books, notebooks, tutor personas, and other activities. The platform supports multi-agent problem solving, retrieval-augmented learning with source-traced and page-level citations, math animations, interactive visualizations, and tutor bots built from personal study materials. Its knowledge sources include personal documents, EPUBs and books with annotations, GitHub repositories, web search, and connected libraries; documented retrieval and ingestion options include GraphRAG, PageIndex, LightRAG, linked knowledge bases, Obsidian, and configurable parsing and vector backends. DeepTutor also includes courses, research workflows, an ecosystem of MCP services and tool or capability plugins, and integrations with connected coding agents and other partners. It can run locally or through Docker.
deja-vu is a local memory and search layer for AI coding agents. Its Go binary indexes session histories already stored on a machine by Claude Code, Codex, Cursor, and other supported agents, including sessions recorded before installation, without using an LLM or embeddings. It provides natural-language and direct search, session inspection, context retrieval, JSON output, redaction, usage statistics, and MCP-based cross-agent recall; the index strips keys and tokens as it is built and can synchronize append-only memory between machines over SSH. Recall can be injected automatically at session start and before prompts, file edits, or commands, and a post-command hook can retrieve what followed a matching failure when the agent supports those hooks. It also indexes files opened during each turn, executed commands and exit statuses, and exact spans replaced by edits, rather than only conversation text. Decisions can be marked as accepted or rejected with notes, and results can report when indexed files have changed since a session. The repository reports sub-millisecond lookups over 5 GB of history, with benchmark harnesses for LongMemEval-S and LoCoMo. The project is distributed as a local binary with installation options including a shell script, Homebrew, Go, npm, Scoop, release archives, and agent-specific plugin or MCP bundles. Its optional installation wiring configures supported agents for MCP and session-start recall.
Firecrawl is an open-source web context API and hosted service for AI agents and applications. It searches the web, scrapes individual pages, crawls websites, maps site URLs, and batch-processes large URL sets, returning content as clean Markdown, HTML, screenshots, structured JSON, and other extracted data. It handles JavaScript-heavy pages, rotating proxies, orchestration, rate limits, and blocked content, and can parse web-hosted PDFs and DOCX files. Its interaction endpoint lets users or agents click, scroll, write, wait, and press on a page before extracting content; its agent endpoint gathers web data from a natural-language request without requiring URLs. Firecrawl provides Python and Node.js SDKs, cURL and CLI interfaces, and an MCP connection for AI agents and applications.
i-have-adhd is an open-source skill/plugin for coding assistants that formats responses as ADHD-friendly output. It leads with the next action, numbers multi-step tasks, suppresses tangents, restates state, gives specific time estimates, uses matter-of-fact errors, limits lists to five items, and ends with one concrete next step. The skill is installed through a coding assistant's CLI and can be customized by editing its SKILL.md file. It is licensed under the MIT License and does not require an ADHD diagnosis.
img2threejs is an agent skill that reconstructs an object or character from a reference image as a code-only, procedural Three.js model. It generates a TypeScript THREE.Group factory using primitives, procedural shaders, and generated geometry rather than photogrammetry, mesh extraction, or downloaded art assets. The generated scene includes runtime structures such as pivots, sockets, and colliders for animation, and the project documents separate hard-surface and anatomy-aware reconstruction paths. It runs under Claude Code, Codex, or OpenCode and provides live browser demos whose models can be inspected as generated source.
jcode is an open-source, Rust-based terminal harness for AI coding agents and LLM workflows, designed for interactive development across multiple sessions. It supports resumable sessions, session search, multi-agent swarm and helper-agent workflows, compatible model-provider integrations, MCP, browser automation, and a self-development mode; installation scripts target macOS, Linux, and Windows. For automatic memory, jcode embeds turns and responses as semantic vectors, searches a graph of stored memories using cosine similarity, and feeds relevant results into the conversation. A memory sideagent can verify and expand retrieval, while periodic extraction stores new memories and ambient consolidation reorganizes entries and checks for staleness and conflicts. Explicit memory tools and traditional retrieval over previous sessions are also available. Its terminal UI includes side panels, diff views, inline Mermaid diagrams, information widgets, custom scrolling, and real-time rendering. The project supplies a Mermaid renderer without browser or TypeScript dependencies and publishes benchmarks covering RAM consumption, startup and input-readiness times, and memory scaling across multiple active sessions.
Kandev is an open-source, self-hostable AI Kanban and development environment for orchestrating coding agents and reviewing their work. It organizes tasks in Kanban and pipeline views, supports parallel execution and multi-step agentic workflows, and isolates concurrent work with Git worktrees. Its integrated workspace combines a file editor, file tree, terminal, browser preview, chat, and Git changes for reviewing and iterating on agent output. Agents can run through local, Docker, SSH, or cloud runtimes, with support for multiple agent providers; the project also provides configurable workflows, agent profiles, prompts, review gates, scheduled or webhook-triggered automations, and no telemetry.
Kimi Code CLI is an AI coding agent developed by Moonshot AI that runs in a terminal. It reads and edits code, runs shell commands, searches files, fetches web pages, and selects subsequent actions based on the feedback from those operations. It works with Moonshot AI's Kimi models and can be configured with other compatible model providers. The tool is distributed as a single binary and provides an interactive terminal UI. It accepts video input, supports conversational configuration of Model Context Protocol (MCP) servers, and can install skills, MCP servers, and data sources from its marketplace or GitHub repositories. Built-in coder, explore, and plan subagents can work in isolated contexts, while lifecycle hooks can run local commands at selected points in a session. Kimi Code CLI also supports the Agent Client Protocol (ACP), allowing compatible editors and IDEs such as Zed and JetBrains to drive sessions through the `kimi acp` command. The repository documents installation for macOS, Linux, and Windows and provides OAuth or Moonshot AI Open Platform API-key login options.
LikeC4 is an open-source modeling language and toolset for describing software architecture as code and generating live, browsable diagrams from the model. It is inspired by the C4 Model and Structurizr DSL, while allowing users to define custom notation, element types, and nested architecture levels. The project provides a CLI for previewing diagrams, along with a VS Code extension and other tools for working with the models. LikeC4 is distributed under the MIT License.
npcpy is a Python library for research and development with multimodal language models, agentic AI, and knowledge graphs. It provides primitives for defining personas, making direct language-model calls, creating tool-using agents and multi-agent teams, and building AI applications with local providers such as Ollama, llama.cpp, omlx, and LM Studio as well as cloud providers. Its NPC Context-Agent-Tool data layer is designed to enforce context and tool-use rules through software rather than prompts. The Agent class includes tools such as shell execution, Python, file editing, and web search, while ToolAgent supports custom tools; the examples include image generation, Hugging Face image-dataset retrieval, and diffusion-model fine-tuning. The library is distributed through PyPI.
Pi is an open-source AI agent harness and toolkit from earendil-works for building and running coding agents. Its packages provide a unified multi-provider LLM API, an agent runtime with tool calling and state management, an interactive coding-agent CLI, a terminal UI library with differential rendering, and vendor-neutral telemetry contracts and adapters. Slack and chat automation are provided through a separate package project. Pi does not provide built-in restrictions for filesystem, process, network, or credential access; it runs with the permissions of its launching user and process. The project documents containerization and sandboxing approaches, including a local Linux micro-VM, Docker, and policy-controlled sandboxing, for stronger isolation.
Pi Web is an open-source local browser interface for the Pi coding agent, developed in the agegr/pi-web repository. It uses Pi’s local configuration and session files, allowing users to browse, resume, rename, export, delete, and branch project-grouped conversations; run agent turns; and inspect running state, context usage, costs, and compaction details. New sessions create independent session files, while “Edit from here” creates a branch within the current session. Its project workspace supports file browsing and uploads, Git diff inspection, automatic previews for source files, Markdown, images, audio, PDFs, and DOCX files, and Git worktree switching. Configuration panels manage provider logins and API keys, models and model tests, plugin packages, and skills. The interface supports English, Simplified Chinese, and Traditional Chinese, follows the browser language initially, and includes a language switcher. Pi Web is distributed through the `@agegr/pi-web` npm package, requires Node.js 22.19.0 or newer, and listens on `127.0.0.1` by default. Remote binding and HTTP Basic Authentication are available, but the documentation warns that Basic Auth does not encrypt credentials in transit and that the service should not be exposed over plain HTTP without a trusted reverse proxy or VPN. It reads Pi agent data from `~/.pi/agent` by default, shares Pi’s model, settings, and credential storage, and limits file browsing to known project or session roots rather than providing general filesystem access.
Playwright is an open-source framework for web testing and browser automation, developed in the Microsoft repository. It drives Chromium, Firefox, and WebKit through a single API and includes an end-to-end test runner with isolated browser contexts, auto-waiting, web-first assertions, resilient locators, parallel execution, and reusable authentication state. Its tooling also includes a browser-automation library, CLI for coding agents, MCP server for AI-agent and LLM-driven automation, and tracing that records actions, DOM snapshots, network requests, console messages, screenshots, and videos for debugging.
PortalJS is an open-source, AI-native framework maintained by Datopian and its community for building data portals. Its agentic skills advise on storage, compute, catalog, access, hosting, and metadata, then scaffold the result as plain, editable Next.js code. Claude Code skills can create a portal, add CSV or JSON datasets, connect to backends such as CKAN, and deploy it; generated portals include a home page, catalog, and dataset showcase, with support for configurable data providers, charts, maps, and schemas. A bare template is also available without the AI skills, and the project presents Git, object storage, Parquet, and DuckDB as an optional modern architecture while retaining traditional datastores as supported choices.
An open-source collection of common React job-interview questions with detailed answers and explanations in Spanish. The repository organizes material from beginner to intermediate topics, including JSX, components, props and state, hooks such as useState and useEffect, rendering, hydration, server-side rendering, controlled components, custom hooks, and React design patterns. Its companion site, reactjs.wiki, supports searching, saving, and sharing questions.
Strix is an open-source AI penetration-testing tool from usestrix that uses autonomous, multi-agent AI pentesters to dynamically run applications, perform reconnaissance and exploitation, and validate vulnerabilities with working proof-of-concept exploits rather than static-analysis findings. Its developer-oriented CLI produces remediation guidance, generated patches, and pentest reports, and it can run scans in CI/CD pipelines to detect insecure changes before deployment. The tool runs in a Docker sandbox and requires an LLM API key; the associated Strix platform supports repository and domain testing, continuous scanning, and integrations with development and issue-tracking workflows.
Temporal is a durable execution platform developed by Temporal Technologies for building scalable applications and workflows. Its server executes application logic as Workflows, automatically handling intermittent failures and retrying failed operations so applications do not need to implement all timeout, retry, and recovery logic themselves. Developers implement Workflows, Activities, and Workers with SDKs for multiple programming languages, then use the Temporal server, CLI, and Web UI to run and inspect them. The open-source Temporal server originated as a fork of Uber's Cadence and is distributed under the MIT License.
USBridge Remote is a unified client for managing remote machines, combining software-based remote desktop control with integration for USBridge KVM hardware. Its client runs on Windows, macOS, Linux, Android, iOS, and the web, while an agent on the target machine handles screen capture, input injection, and Tailscale networking. The system provides live remote desktop, virtual device passthrough, snapshot management, shared clipboard transfers, encrypted peer-to-peer connections, and Moonlight-based low-latency streaming. It is distributed as beta software, is free without session or connection limits, and does not require an account on the target machine; the web client has feature and performance limitations compared with native applications.
Searchable transcript of Top Dev Tool Projects : Playwright, Firecrawl, Temporal, LikeC4, Kimi Code CLI & Strix — ManuAGI - AutoGPT Tutorials (20:57). 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 dev tools, and this weekly dev tool project update video brings them together in one place. Top trending open source and best dev tool projects this week cover terminal coding agents, local code knowledge graphs, web automation, speech recognition, and self-hosted platforms. You'll discover useful and trending developer tools you can start using right away, explained fast and to the point.
00:24 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 cutting-edge agent frameworks, make sure to check it out.
00:54 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, PiWeb, local browser workspace for the Pi coding agent. PiWeb is a local web interface for the Pi coding agent that turns terminal sessions into a readable browser workspace.
01:17 It reads your local Pi session files and lays them out with structured tool calls, clean markdown, and project navigation beside the chat, so you pick up past work without digging through terminal history. You browse earlier conversations by project, continue from an old message, or fork a session into a separate route to try a different direction safely.
01:36 A file panel previews source, docs, images, audio, and PDFs while the agent works, and the web UI handles models, API keys, and skills. Built with TypeScript and Next.js, it runs through NPX and stays on your machine. It suits developers using Pi day to day. Run it and open your sessions. Nifcan Quono de day Ryan. Project number two, like C4 architecture as code with live generated diagrams.
02:05 Like C4 is a modeling language and tool set for describing software architecture as code and generating diagrams from that model. It solves a familiar problem. Hand-drawn architecture diagrams drift out of date the moment code changes. With like C4 you write the model in a text file and the tools render it into live browsable diagrams that stay in sync with the source.
02:25 It draws on the C4 model and structurizer DSL, but lets you define your own notation, element types, and any number of nested levels so the view fits your system. Developers run its CLI with NPX like C4 start to preview, use a VS Code extension while editing, and export views for sharing. Written in TypeScript, it suits teams documenting evolving systems.
02:49 Model your architecture and watch it render. Project number three, Code Review Graph, local code knowledge graph that trims AI context. Code Review Graph is a command-line tool and MCP server that builds a persistent structural map of your code base so AI coding assistants read only the files that matter. It addresses a real cost. Most AI tools reread the whole project on every task, wasting tokens.
03:15 It parses your repository with tree-sitter into a graph of functions, classes, imports, and their call and inheritance edges stored locally in SQLite. When a file changes, blast radius analysis traces every caller, dependent, and test affected so the assistant sees a minimal review set. It supports 24 languages plus notebooks, updates incrementally on each save or commit, and installs across Claude, Cursor, CodeX, and many more with one command.
03:43 It suits developers running large or monorepo projects. Install it and map your code. Project number four, Kim iCode CLI, terminal AI coding agent from Moonshot. Kimiko's CLI is an AI coding agent from Moonshot AI that runs in your terminal. It reads and edits code, runs shell commands, searches files, fetches web pages, and picks its next step from the feedback it gets.
04:06 So, a plain language request turns into real work in your project. It ships with Moonshot's Kimiko's models and can be pointed at other compatible providers. It installs as a single binary with one command and no node.js setup. Then starts an interactive TUI you log into through OAuth or an API key. You drop a screen recording into the chat as video input, configure MCP servers conversationally, install skills and plugins from a marketplace, dispatch coder, explore and plan sub-agents, and drive sessions from Zed or
04:39 JetBrains over the agent client protocol. It suits developers who work from the terminal. Install it and start a task. Project number five, JCode, fast rust terminal harness for multi-session agents. JCode is a terminal coding agent harness written in rust, built for multi-session workflows, deep customization, and low resource use. It gives the agent a human-like memory.
05:02 Each turn is embedded as a vector and relevant past entries are recalled automatically, so context returns without burning tokens on manual lookups. A swarm mode runs several agents in one repository with the server notifying them when files shift and routing messages between them, so conflicts resolve on their own. It logs into many providers through OAuth or API keys, including Claude, OpenAI, Gemini, and Copilot, and can resume sessions started in Codex, Claude Code, Open Code, or Pi.
05:34 A side panel renders mermaid diagrams, a built-in browser tool drives Firefox, and a self-dev mode lets it edit and rebuild its own source. It suits developers running many parallel sessions. Install it and start a session. Project number six, Deep Tutor, agent native personalized learning assistant you self-host. Deep Tutor is a self-hosted agent native tutoring platform that turns your own study materials into a personalized learning system.
06:02 It brings six modes into one thread, so a chat question can escalate into multi-agent problem-solving, quiz generation, deep research, math animation, or interactive visuals without losing context. You upload PDFs, office files, and markdown to build rag-ready knowledge bases. Then a persistent memory tracks what you study and how you learn. It compiles interactive living books, offers a multi-document AI co-writer, and runs autonomous tutor bots with their own memory and personality across channels like Telegram and
06:35 Discord. Built with Python and Next.js, it works with many LLM providers and installs through a guided setup, manual install, or Docker. With a full CLI for humans and agents, it suits learners and educators. Set it up and start learning. Project number seven, Pi Agent Harness, AI agent toolkit with a self-extensible coding agent. Pi Agent Harness is an AI agent toolkit and a self-extensible coding agent for the terminal.
07:03 It splits the work of building agents into clean packages, so developers don't rebuild the same plumbing each time. A unified LLM API talks to OpenAI, Anthropic, Google, and other providers through one interface. An agent runtime handles tool calling and state. A terminal UI library renders the interface with differential updates, and an interactive coding agent CLI ties them together.
07:26 The agent can explain itself when asked, and separate packages cover Slack and chat automation. Pi ships no built-in permission system, so the docs describe containerizing it with a Linux micro VM, plain Docker, or a policy-controlled sandbox for stronger boundaries. Written in TypeScript and distributed on NPM, it suits developers building or running agents.
07:50 Install it and start a session. Project number eight, Strix, autonomous AI penetration testing agents for apps. Strix is an open-source penetration testing tool that runs autonomous AI agents to find and validate vulnerabilities in your applications. It targets the gap between slow manual pen tests and static scanners that flag false positives. Strix runs your code dynamically and proves each finding with a working exploit.
08:15 The agents carry a full offensive toolkit, including an HTTP interception proxy, a browser for testing XSS and CSRF, a shell, and a Python sandbox for writing proof-of-concept exploits. A graph of specialized agents handles reconnaissance, exploitation, and post-exploitation together, and the CLI reports findings with remediation guidance and can generate patches.
08:38 It works with providers like OpenAI, Anthropic, and Google, run scans in a Docker sandbox, and slots into CI to check pull requests. It suits developers and security teams. Point it at a target and start testing. Project number nine, Firecrawl, web API turning any site into agent-ready data. Firecrawl is an open-source API for searching, scraping, and interacting with the web at scale, built to feed clean data to AI agents and apps.
09:08 It solves the messy parts of web data. It renders JavaScript heavy pages, handles rotating proxies, rate limits, and orchestration, then returns clean markdown, structured JSON, or screenshots. Its core endpoints search the web with full page content, convert any URL into LLM-ready formats, and let you click, scroll, and type on a page before extracting.
09:30 Extra endpoints crawl whole sites, map their URLs, batch scrape thousands at once, and run an agent that gathers data from a plain language prompt. It ships SDKs for Python, Node, Go, Java, and more, connects to agents through a CLI skill and MCP, and can be self-hosted. It suits developers building AI apps that need live web data. Get a key and start scraping.
09:55 Project number 10, Temporal, durable execution platform for reliable workflows. Temporal is a durable execution platform that lets developers build scalable, reliable applications without giving up productivity. This repository holds the Temporal server, which runs units of application logic called workflows, and keeps them resilient by automatically handling intermittent failures and retrying failed operations.
10:20 That removes the usual burden of writing custom retry, timeout, and recovery logic. For long-running or distributed processes, you write workflows, activities, and workers in a supported language, then run them against the server. Inspecting execution through a command line tool and a web UI. Written in Go, it started as a fork of Uber's Cadence and is built by the creators of that project.
10:43 It installs locally through a single dev server command for quick testing. It suits back-end developers building resilient services. Start the server and run a workflow. Project number 11, EMG23JS. Rebuild a reference image as procedural 3JS code. EMG23JS is an agent skill that rebuilds the object in a reference image as a code-only procedural 3JS model.
11:07 It is reconstruction by code, not photogrammetry or mesh extraction. From one image, it writes a TypeScript factory that recreates the object from primitives, procedural shaders, and generated geometry with a runtime hierarchy of pivots, sockets, and colliders, so the result is ready to animate. It is built to save tokens by pushing mechanical work into deterministic Python scripts that use only the standard library, reserving the model's attention for visual judgement, a staged pipeline moves from blockout to
11:40 material, surface, and lighting passes, and each pass unlocks only after a side-by-side render passes a vision check. It runs under Claude code, Codex, or open code. It suits developers building 3D scenes with agents. Install it and rebuild an image. Project number 12, Nenp cpy, Python library for building multimodal AI agents. Nenp cpy is a Python library that gives developers the building blocks for research and development with multimodal language models, agents, and knowledge graphs.
12:11 It centers on a context agent tool data layer, so you define personas called NPCs with directives and tools, then compose them into multi-agent teams led by a coordinator. Ready-made agent classes cover default tools, custom tools with MCP, and auto-executing code, and templates called Jinxes chain multi-step prompt pipelines. It builds and evolves knowledge graphs from text through waking, sleeping, and dreaming phases, generates images, audio, and video, and supports fine-tuning with SFT and reinforcement learning.
12:44 It works with local runtimes like Ollama and Llama, CPP, and cloud providers through Little M, and installs with pip. It suits researchers and developers building AI applications. Install it and create an agent. Project number 13, I have add HD, agent skill for direct action-first coding output. I have add HD is a skill for coding assistance that stops them from burying the answer.
13:09 It fixes a common frustration. Agents open with filler like, "Great question," wander through context, and close with, "Hope this helps," leaving you to scroll for the actual step. This skill enforces 10 rules that reshape output, so the reply leads with the next action, numbers multi-step tasks, restates the current state each turn, gives specific time estimates, caps lists at five items, and drops preambles, recaps, and closers.
13:37 It installs through the plugin marketplace for Claude code or Codex, works with several other agents, and can be set to apply on every session. You can fork it and edit the rules in skill.md to fit your taste. It suits developers who want terse, scannable answers. Install it and get straight to the point. Project number 14, Preguntas Entrevista React, React interview questions with answers in Spanish.
14:03 Preguntas Entrevista React is an open-source collection of common React interview questions answered and explained in Spanish. It helps developers prepare for technical interviews by gathering the topics that come up most in hiring processes, from beginner concepts to more advanced ones, each with a clear explanation and practical code examples. Rather than memorizing definitions, you practice with reasoning behind every answer, covering areas like use effect, fetch cancellation, and hydration.
14:29 The questions live in the repository as markdown, and a companion website built with Next.js lets you search them, save favorites, and share individual questions. It suits Spanish-speaking developers getting ready for React roles. Browse the questions and start preparing. Project number 15, Nana's USB Bridge Remote, unified client for low-latency remote machine control.
14:51 USB Bridge Remote is a client for managing remote machines that brings hardware BIOS level access and software remote desktop into one interface. It removes the usual split between separate tools by letting you handle USB Bridge KVM devices and software agents from a single dashboard. Add a machine, connect, and you are in. It splits into two parts, a client that runs on your workstation, laptop, or phone to control sessions, and an agent that runs on the target machine to handle screen capture, input injection, and
15:24 networking. Video streams at up to 2K through native Moonlight integration. Built-in Tailscale gives encrypted peer-to-peer tunneling without port forwarding, and the Linux agent supports Wayland without permission prompts. Written mostly in Go, it runs on Windows, macOS, Linux, Android, and iOS. It suits people managing remote computers. Install it and connect a machine.
15:47 Project number 16, Portal JS, AI native framework for building data portals. Portal JS is an open-source AI native framework for building data portals. With the help of a coding agent, it tackles a task that is more than a website, deciding where data lives, how it is versioned, searched, served, and governed, then wiring a front end on top. You describe the portal you want, and Claude code skills both advise on an architecture and scaffold it as plain editable next JS code with no lock-in.
16:17 Every portal is built from three surfaces, a home page, a catalog, and a per data set showcase, which read data through one data provider contract, so the back end can be static files, CKAN, GitHub, or a lakehouse without touching a page. Skills add data sets, charts, maps, schemas, and deployment. It suits data teams publishing open data. Create a portal and load your data.
16:43 Project number 17, CanDev, self-hostable Kanban environment for orchestrating coding agents. CanDev is a self-hostable Kanban and development environment for running coding agents in parallel and reviewing their work. It fills a gap in terminal agent tools. They run agents well, but reviewing and iterating on changes there does not scale. CanDev organizes work across Kanban and pipeline views, executes many tasks at once, and assigns agents from any provider, then gathers the output in one workspace with a file editor,
17:15 tree, terminal, browser preview, and get changes. Agentic workflows chain different agents per step. Get work trees keep parallel agents from conflicting, and runtimes can be local, Docker, SSH, or cloud. Written in Go and Next.js, it runs on your own infrastructure with no telemetry and is reachable from anywhere over a VPN. It suits developers coordinating agent fleets.
17:40 Install it and orchestrate your agents. Project number 18, Deja Vu, local memory layer searching your coding agent logs. Deja Vu is a command line memory layer for coding agents that makes their past work searchable. It solves a quiet waste, Claude Code, Codex, and Open Code write every conversation to local files. Gigabytes of solved problems you cannot search, so agents re-debug what they already fixed.
18:06 Deja is a single zero dependency binary that turns those histories into a local index. You search it directly or wire in an MCP recall tool, so the agent answers, "We fixed this weeks ago." and an auto recall hook can feed relevant memory into each session before you ask. It redacts API keys and tokens at index time, reports usage stats, shares sanitized session digests, and syncs memory between machines.
18:31 Written in Go, nothing leaves your machine. It suits developers who reuse past agent work. Install it and search your history. Project number 19, Dans, Adar ASR V1, Arabic first generative speech recognition models. Dynan Sarid Zar V1 is a family of Arabic first speech recognition models from Adar AI, and this repository is the developer hub for using them.
18:57 It treats transcription as audio-conditioned next token prediction over a text vocabulary using a language model decoder rather than a CTC or transducer objective, and it is trained on more than 300,000 hours of labeled Arabic audio. It transcribes modern standard Arabic and major dialects like Gulf, Egyptian, Levantine, and Maghrebi, plus code-switched Arabic English and English for 30 languages in total.
19:22 Two tiers share one architecture, a 0.78B flash model for real-time and on-device use, and a 2.35B turbo model for accuracy. The repo provides model pointers, benchmarks, and copy-paste inference through transformers and llama.cpp gguf with weights on huggingface. It suits developers building Arabic voice applications. Download a model and start transcribing.
19:49 Project number 20, Playwright, one API to drive Chromium, Firefox, WebKit. Playwright is a framework for web testing and automation that drives Chromium, Firefox, and WebKit through a single API. It solves the pain of flaky, browser-specific end-to-end tests by giving you one consistent way to script real browsers across every engine. Its test runner isolates each test in a fresh browser context, auto-waits for elements to be actionable, and retries web first assertions, so timing bugs mostly disappear, while resilient
20:22 role and label locators mirror how users see the page. Traces capture screenshots, DOM snapshots, and network calls for debugging failures. Beyond tests, it ships as a library for scraping PDFs and screenshots, plus a CLI and MCP server, so coding agents browsers. It runs on Linux, macOS, and Windows with bindings for TypeScript, Python, .NET, and Java. It suits developers testing web apps. Install it and write your first test. Thanks for watching. See you in the next update.