Agent Console is an open-source, local-first observability tool by LockedIn Labs for AI coding agents. It reads Claude Code and Codex transcripts to show token usage by cache reads, cache writes, output and uncached input, model and session activity, list-price cost estimates, burn rate and dollars per hour. A self-hosted hub can combine privacy-limited usage metadata from multiple machines over TLS, while keeping prompts, replies, file paths and file contents local. It also provides optional Claude Code OpenTelemetry and Kong or LiteLLM metric ingestion, a Prometheus endpoint, Grafana dashboard, presenting mode, alerts, and policy hooks. It runs locally from source or release downloads on macOS, Linux and Windows, requires Node.js 22 or newer for the source route, and is licensed under the MIT License.
Agent Office is a self-hosted 3D workspace for coordinating coding agents across GitHub repositories. It creates a floor for each project, places Claude Code, Codex, OpenCode and other supported agent workers at desks, streams their terminals, and lets users assign issues, queue tasks, provide Git worktrees, inspect changes and open pull requests. Workers can also list, hire, message and dismiss other workers through the agent-office MCP server or command-line interface. The office includes shared chat, voice, screen sharing and a whiteboard, plus a 2D /lite interface for phones and slower computers. It runs locally or can be deployed to AWS, Azure, Railway, Fly.io, Dokploy or an Ubuntu/Debian server, with multi-user accounts and per-user Claude and GitHub sign-ins. The repository describes it as work in progress, subject to breaking changes, and distributes it under the MIT license.
Antigravity Fixer is an open-source command-line tool for diagnosing and fixing Google Antigravity, Gemini Code Assist, and related 9router eligibility and login errors. It checks account-related conditions, removes stale credentials and caches, terminates stuck processes, and can revoke OAuth connections before guiding the user through Google sign-in and age verification. It supports the Antigravity CLI and desktop/IDE installations on their platform-specific credential stores and application paths, and includes a clean-only mode for local cleanup. The tool is written in Python, uses Camoufox for headless browser automation and Rich for its terminal interface, prompts for passwords at runtime without writing them to disk or logs, and is distributed under the MIT license.
Arc Library is an open-source React component and block library by Elia Kuratli, providing interfaces with calm physical motion, keyboard support, and a reduced-motion path for every animation. Components use CSS modules with shared design and motion tokens, and can be installed through the shadcn CLI or copied into a project as source code; there is no runtime package to keep synchronized. It supports Next.js and Vite with or without Tailwind, and includes controls, inputs, overlays, feedback elements, data visualizations, tables, media components, text effects, and complete interface blocks. The free repository is MIT licensed. Components use Motion for animation and may use lucide-react or Radix UI primitives; theming is controlled through semantic CSS variables with separate light and dark values. Arc also provides an MCP server and machine-readable documentation for AI coding tools, while paid Arc Pro components are distributed separately.
Blueprint Animation is a Claude Design skill for presenting UX redesigns and design rationale as continuous blueprint animations. In Redesign mode, it turns the Before screen into a blueprint, animates only the changed parts, and reveals the After screen; in Explain mode, the unchanged screen is converted into a blueprint module by module with guides, dimensions, and labels. Each presentation uses 3–6 numbered steps containing a focus stage, blueprint conversion, construction or annotation, UI reveal, and an explanation of the fix or design decision. The skill reproduces Figma designs using their values, fonts, and exported icons, checks built screens against the Figma export, and includes rules for motion, text placement, UI swaps, performance, and handoff QA. It is distributed as a ZIP upload for Claude Design and is licensed by Oğuz under CC BY-NC 4.0; commercial use requires written permission.
CalcKernel is a statically typed language and compiler for numerical computation kernels: focused calculations that can be called from larger applications. Its typed functions, structures, integer and floating-point values, conditions, loops, and caller-owned data slices define an explicit boundary between calculation code and application responsibilities such as user interfaces, input/output, and data storage. The `ckc` compiler can check, run, and build `.ck` files into native programs or libraries, C source, or WebAssembly modules; native and C builds can optionally enable integer-overflow and slice-bounds checks, while WebAssembly builds support selected bulk-memory and SIMD optimizations. The project includes a runtime, tests, examples, documentation, public releases, and an optional Visual Studio Code extension, and is distributed under the MIT License.
Claude is an AI assistant developed by Anthropic, positioned as 'The AI for Problem Solvers'. It is a general-purpose AI system used for tasks such as generating code prompts, refining requirements, creating advertising strategy and copy, and processing creative content like storyboarding and video prompts.
A Claude Code skill for creating motion-design videos in code using HTML, Playwright, ffmpeg, Python, and Chromium. It computes each frame from a deterministic seek(t) function, renders frames with Playwright, blends subframes for motion blur, and encodes the result with ffmpeg. Its motion system includes closed-form springs, log-space camera keyframes, shared-element handoffs, masked word-by-word text, color floods, and ports of selected 21st.dev components. The workflow moves from inputs and a beat map through still-frame approval, full rendering, pop detection, audio generation, and final muxing. Audio processing finds musical drops by band energy, aligns the song to a key visual frame, places sound effects at measured peaks, and applies two-pass loudness normalization. The repository also includes a remake mode for frame-locked recreations with shot analysis, side-by-side verification, and royalty-free replacement audio. The code is MIT-licensed; downloaded third-party media remains under its respective license.
Coucou is an open-source desktop companion for Claude Code and other coding-agent sessions, appearing in the macOS notch or as a top-of-screen island on Windows. It displays session events such as files read, edits, and commands run; surfaces Claude Code permission requests for Allow/Deny approval; opens the relevant terminal on macOS; provides built-in Claude chat; and accepts dropped files as context. It can also attach a macOS window as Claude context and display optional integration pills for services including GitHub, Vercel, Notion, and others. On macOS, a hook script forwards Claude Code events over a Unix socket and waits for approval responses, while a borderless NSPanel and a SwiftUI Canvas render the animated Mochi character. The native Swift 6, SwiftUI, and AppKit application has no third-party dependencies. On Windows, a Tauri 2 application uses Rust, TypeScript, a transparent always-on-top window, Canvas 2D rendering, and a named pipe for hook events. The project states that it collects no telemetry, requires no account, and stores credentials in the macOS Keychain or Windows Credential Manager. Coucou is distributed from the project's releases and can also be built from source. The code is licensed under MIT; the Mochi name, character, icon, sounds, and media are separately reserved by the author. The Windows installer is temporarily unavailable according to the README because its unsigned build was flagged by Microsoft Defender.
Deck as Controller is a Decky Loader plugin that turns a Steam Deck into a wireless Bluetooth gamepad for a Mac or iPad. It can present the Deck as a PS5 DualSense, PS5 Edge, or Xbox Elite Series 2 controller, translating input from the Deck's built-in controller and supporting trackpads, gyro on PS5 profiles, selected back buttons, rumble handling, battery reporting for PS5 profiles, automatic reconnection, and screen-off operation. The plugin runs a daemon using the Deck's system Python: it temporarily reconfigures BlueZ to free the HID channels and serve the selected Bluetooth HID profile, takes exclusive access to the built-in controller over USB, and translates its reports into the selected controller format. Turning the plugin off or rebooting restores the stock Bluetooth setup and Deck controls. It is an early beta build with reported latency and connection issues, and is released under the MIT License.
Dioramas is a free, open-source framework for building cinematic, interactive 3D websites in which the 3D scene forms the page. Its asset pipeline uses Nano Banana 2 on fal to generate product images and Meshy 7.1 on fal to convert them into textured 3D models, then optimizes the models with meshopt compression and WebP textures. The Three.js engine provides HDR post-processing, ambient occlusion, bloom, tone mapping, volumetric spotlights, planar reflections, procedural textures, and Lenis/GSAP scroll choreography, along with cursor, preloader, and navigation components. The repository includes 20 complete example sites, their asset prompts, an asset manifest, and scripts for generation, optimization, screenshots, and review. It runs locally with npm, ships its models in the repository, recommends a desktop GPU, and is licensed under MIT for code and CC BY 4.0 for generated models and images.
DroidDeck is an Android application that brings a SteamOS-like gaming interface to supported Adreno handhelds. It runs Valve's Steam client in Big Picture mode and uses ARM64 Proton to run Windows games; an optional Linux desktop provides Flatpak applications and games from Flathub, along with emulators. It requires Android 9 or newer and a supported Adreno device, does not require root, and is distributed as an APK with a separate Linux runtime. Compatibility and performance vary by device; desktop compositing uses software rendering, and Firefox sandboxing is reduced under proot. The project is licensed under GPL-3.0 and is not affiliated with Valve.
DSCODE is a macOS terminal coding agent and CLI built on DeepSeek Harness. Its persistent shell retains the working directory, environment, background jobs, edits, searches, patches, and test runs, while the TUI, CLI, and scripts share one session runtime. Sessions can hand tasks to other sessions, run side questions in read-only child sessions, delegate work to child agents in isolated Git worktrees, and use an independent read-only model to review changes and approve permissions. The `dscode exec` command supports non-interactive scripts and CI, with output streamed to stdout and tool activity and session identifiers sent to stderr. It also supports MCP servers, triggers, cross-session memory, multiple model providers, and a workspace-write sandbox with human-approval and automatic-review modes. Distribution is available through npm, Homebrew, GitHub release tarballs, and source; the project requires macOS 14 or later and is licensed under MIT. It is an independent community project, not an official DeepSeek product.
GPUI Fast is an experimental fork of GPUI, the UI framework used by the Zed editor, focused on retained-mode rendering and native view composition. Its retained layout engine preserves views whose inputs have not changed and reuses unchanged layout nodes, so subsequent frames render, lay out, prepaint, and paint only affected parts while replaying the previous output for retained views. Changes in entities, globals, list or scroll state, hover, bounds, content masks, and window refreshes invalidate the relevant retained state; retention can also be disabled for comparison and debugging. The project also explores placing native views such as a WebView inside a GPUI window while keeping GPUI popovers, menus, and dialogs above them. It aims to preserve GPUI's existing API and keep its changes in separate fast/ directories so applications can adopt the experimental behavior without rewriting their UI code. Applications can use the repository as a Git dependency in place of upstream GPUI; code that reads state outside GPUI's observable entities may require an explicit cx.notify() when that state changes. GPUI Fast is intended for experimentation and measurement, and its internals may change. It is based on upstream GPUI from Zed and is licensed under Apache-2.0.
hntui is a terminal user interface for browsing Hacker News. It retrieves data from the public Hacker News Firebase API and provides the Top, New, Best, Ask, Show, Jobs, and Saved feeds, with story comments, article links, and browser opening. Navigation supports Vim-style keys and mouse input; comment threads can be collapsed, posts can be saved locally in a JSON file, and the interface includes dark and light themes. It runs in modern true-color terminals with UTF-8 and mouse support. The project distributes dependency-free binaries for macOS and Linux and can also be installed or run through Bun and npm. It is released under the MIT license.
hypoarena is a fully offline scientific hypothesis-discovery workbench and command-line tool. It generates deterministic synthetic literature containing planted causal chains, competing hypotheses, paraphrase clusters, distractors, and contradictions, then runs a generate–debate–evolve loop through pluggable agent adapters that propose, critique, and revise hypotheses. Its pipeline includes span-level citation and claim-grounding checks, paraphrase deduplication using hashing, n-gram Jaccard, TF-IDF, and MinHash LSH, Bradley–Terry/Elo tournaments for ranking hypotheses or agents, hypothesis evolution operators, and Bayesian evidence accumulation. It produces Markdown and self-contained HTML reports with rankings, grounding flags, deduplication clusters, belief updates, run metadata, and limitations. The core runtime depends on NumPy; an optional CPU-only PyTorch extra demonstrates a trainable ranker. Examples and tests run without network access, real model calls, or downloaded weights, and the project is licensed under the MIT License.
iCode is a lightweight, extensible, fully offline terminal development platform and toolkit for orchestrating AI agents and workflows, developed as an open-source project by openJiuwen-ai. Its terminal interface shows the active agent, model, approval mode and session; renders Markdown and diagrams; and provides workflow execution views with per-node models, tool calls and token spending. Agents are configured rather than coded and can include instructions, tools, sub-agents, skills, MCP servers and memory. Model profiles from different providers can be switched during a conversation, while file diffs, turn-based rollback, token and cache usage, model/tool timelines, and an embedded shell expose and control execution. iCode does not include models, telemetry, analytics or crash reporting; it uses only the providers, MCP servers and hooks configured by the user. The repository describes the project as a work in progress, requires uv to run, and distributes it under the Apache License 2.0 alongside separately licensed third-party components.
imajev is an open, locally runnable multimodal typed-decision model and repository maintained by Mohit Garg. It accepts up to two images, application state, and typed questions with user-defined answer options, then returns probabilities for each option together with an unknown probability, confidence, and abstention status. Example tasks include checking a product listing against its photo, comparing a returned item with a reference image, routing support tickets, and evaluating records against policies. Rather than generating an answer, imajev binds each option to a decision code and reads the model's logits at a decision position through a float32 output head. Its vision tower is frozen while LoRA adapters are trained on the language-model projections; the released 2B, 4B, and 9B variants are based on Qwen3.5 models. Unknown is trained as a first-class outcome for missing, contradictory, or out-of-scope evidence, allowing applications to route low-confidence cases to a person. The repository provides a local server and API with MLX and PyTorch paths, calibration files, playground applications, training scripts, and an ImajevBench evaluation suite. The code, adapters, and stated base models are distributed under Apache-2.0. The README describes English-only operation and limits of up to two images, 32 KB of state, eight questions per request, and 254 options; it also cautions that calibration and testing on application-specific data are needed before setting automation thresholds.
Jeeves is a 9B reasoning Jev-style decision model for producing calibrated probabilities for yes/no, multiple-choice, and rating questions. It can support application decisions such as ticket routing, urgency flagging, and customer-frustration scoring through a Jev-compatible API. The model is based on Qwen3.5-9B with LoRA and a pointer head. It uses supervised fine-tuning and CISPO reinforcement learning to reason before deciding, then scores answer options with a pointer head and converts the scores into probabilities. A block-4 diffusion drafter accelerates reasoning-chain generation. The repository includes training and evaluation code, data-preparation scripts, released model weights, an inference server, CUDA and Metal execution paths, FP8 support, and a Python SDK designed as a drop-in replacement for Jev's SDK. The server accepts structured state and question definitions and returns answers, probabilities, confidence values, and optionally reasoning text.
Job Radar is a self-hosted job-monitoring tool that watches company career boards, detects new and closed postings, scores openings against a resume with an LLM, and sends alerts. It tracks companies across ATS platforms including Greenhouse, Lever, Ashby, Workable, Workday, and Oracle Recruiting Cloud; each job receives a stable ID so new and removed postings can be identified. Its scoring applies code-based seniority and relevance caps in addition to the model's 0–10 score and explanation. It provides daily email and desktop alerts for selected companies, a ranked weekly digest, terminal commands for scanning, alerting, recapping, and listing results, and a local web interface for searching and filtering jobs. The repository requires Python 3.11 or later and a Mac because its scheduled jobs use launchd. It accepts any OpenAI-compatible LLM endpoint, uses Gmail App Passwords for email, and includes an installer that registers daily and weekly schedules. The project is released under the MIT license.
KCC is a cross-platform comic and manga converter for Kindle, Kobo, reMarkable, and other E Ink readers. It processes image folders, comic archives such as CBZ and EPUB, and PDFs, then resizes pages to device profiles, crops margins and page numbers, adjusts contrast and black levels, supports manga and webtoon layouts, and detects or splits two-page spreads. It can produce MOBI/AZW3, EPUB/KEPUB, CBZ, PDF, or image-folder output, with GUI and command-line tools for Windows, macOS, and Linux. The project also includes processing intended to reduce the rainbow effect on Kaleido 3 displays using a 2D discrete Fourier transform. KCC is released under the ISC license and is maintained in the ciromattia/kcc GitHub repository.
Naive-N0.5-Flash is an open-weight mixture-of-experts language model from NaiveAI for coding and AI research. It has 309 billion total parameters, 15.5 billion active parameters, and a native one-million-token context window implemented with a hybrid of 128-token Sliding-Window Attention and DeepSeek Sparse Attention; the latter's indexer scans the history while its backbone attends to the top 2,048 selected tokens, with no full-attention layers. The model is released under the MIT License with inference code, and supports FP8 mixed-precision inference through Transformers on FP8-capable NVIDIA GPUs; the repository reports approximately 315 GB of model weights. NaiveAI also describes NaiveRT, an inference system using mega-kernel fusion, Programmatic Dependent Launch, and speculative decoding.
Omarchy Spaces is a workspace-switcher widget for the Omarchy bar. It displays the application icons on each workspace, expands the active workspace, highlights the focused window, and provides live previews when another workspace is hovered; previews follow the monitor orientation and can be clicked to focus a window. It also supports optional status badges for Claude Code and OpenCode terminals, using hooks or a plugin to report working, waiting, done, and ended states. The widget supports clicking and scrolling to switch workspaces, window-title tooltips, configurable settings saved in Omarchy's shell configuration, and command-line preview and settings controls. It requires Omarchy 4 with the Quickshell bar, Hyprland 0.56 or newer, and jq for the agent hook. The repository is distributed under the MIT License.
onetake is a Claude Agent Skill for creating product launch films, teasers, and feature demos as continuous one-take animations rather than sequences of cuts. It rebuilds product interfaces from screenshots in HTML, places them on the camera plane, and uses a motion library of measured spring, entrance, carry, collision, simulation, camera, and fluid-light movements. Its workflow requires a beat sheet in which each boundary identifies what survives into the next beat, then renders deterministic frames with sampled linear-light motion blur and event-based sound. Before a film is shown to a human, probe.py records frame contents, detects beat boundaries, and uses a continuity oracle to check whether visual elements carry across them. The verification also rejects uniform shot timing, missing stillness, clipped audio, fast movement without blur, and subjects leaving the frame. The repository includes rendering, verification, probing, reference-analysis, still-image, footage, narration, and sound scripts, with draft output at 1080p30 and final output at 4K60. It requires Python with Playwright and Chromium, NumPy, SciPy, Pillow, and Matplotlib, plus FFmpeg and Node; narration additionally uses faster-whisper and Kokoro TTS. It is distributed under the PolyForm Noncommercial 1.0.0 license, which permits personal, educational, research, and other noncommercial use but not commercial use.
OpenAI MCP Extensions is an open-source extension framework for adding ChatGPT-specific capabilities to MCP, allowing developers to build plugins that behave as native ChatGPT features. Supported extension points include sidebar entry points, custom file-extension handlers, composer mentions for searching plugin resources and adding references to messages, and extended forms with thumbnail-based choices. The repository provides TypeScript and Python SDKs for MCP servers and Apps, along with plugin documentation and a specification. It is licensed under Apache License 2.0.
PaperMono Shopping List is an open-source household shopping-list application consisting of firmware for the M5Stack PaperMono e-paper device, a FastAPI and SQLite server, and a phone web UI. The device displays items grouped by aisle, supports touch-based ticking and on-device entry, and works offline by queuing edits on flash. Wi-Fi is enabled only for scheduled synchronization, after edits, or when a stale list is accessed; the phone connects to the server over HTTP. The server maintains item and aisle data and can optionally send new item names to the Claude Code CLI for aisle classification, while manual aisle corrections are retained. The repository is licensed under the GNU GPL v3.0 and describes the project as a personal setup rather than a commercial product; it has been tested only with the PaperMono C153 and the server has no built-in authentication.
QDuo is a macOS menu-bar utility that displays configurable actions at the cursor when text is selected in native apps, browsers, Electron apps, or PDFs. Actions can translate, polish, explain, summarize, search, speak, transform, run shortcuts or shell scripts, or send prompts to DeepSeek, OpenAI, Doubao, Qwen, Ollama, and other OpenAI-compatible endpoints. It can also extract text from a screen-region screenshot. Results can replace the selection, be inserted after it, or be copied to the clipboard. The app supports capsule or ring-shaped popups, restores the clipboard when it must use clipboard access, stores API keys in the macOS Keychain, and uses a JSON configuration file at ~/.config/qduo/config.json. It is distributed as a signed and notarized macOS application through Homebrew or GitHub Releases, requires Accessibility permission, optionally requires Screen Recording for screenshot text, and is licensed under GPL v3.0.
ReelMimic is a local 2D animation tool that uses Claude Code or Codex to create a new video from a reference video and a brief. It analyzes the reference's shot count, pacing, BPM, transitions, framing, colors, and camera movements, then produces a storyboard, character and asset plan, and style frames for approval. After approval, multiple AI agents build different parts in parallel, while separate agents review completed shots and request fixes with before-and-after screenshots. The tool supports seven drawing engines: vector/motion graphics, hand-painted watercolor, crayon picture book, pixel art, paper cut-out stop-motion, whiteboard doodle, and anime cel; new styles are defined in Markdown. It runs on the user's computer, uses the user's Claude Code or Codex account, and does not reuse the reference video's footage, characters, logos, or assets. The repository is MIT-licensed and has been tested primarily on Windows, with macOS and Linux support described as less tested.
Rungic is an AgentOS product that runs alongside Android on compatible phones, providing a Linux desktop with KDE Plasma and desktop applications such as Firefox, Blender, Krita and VS Code. The Linux environment runs in an LXC container on Android's kernel rather than in a virtual machine or emulator, while an Android app connects its display, touch input, sound and camera to the phone, a floating window or a TV. Users can install software through apt, Flatpak or Discover and use the phone as a touchpad and keyboard when casting. Rungic includes a bundled assistant whose reference integration uses a realtime voice model for conversation and Codex for background tasks. The assistant plans tasks, opens applications, clicks, types and returns created files in the conversation; its work can run in a separate desktop workspace so the user can continue using the phone. The system supports multiple agent workspaces, shared project folders, desktop control, phone controls and proactive system-care suggestions that record evidence, investigation results, repair plans and user decisions. Rungic exposes these capabilities through two MCP servers, command-line tools, D-Bus services and standard Linux interfaces, allowing other agents to be adapted to it. Rungic is under active development and in private preview. Its device adaptation currently requires a compatible, unlockable Android phone with an appropriate GKI kernel, ARM64 support and a Snapdragon processor with an Adreno GPU; the repository documents tested Motorola devices and separate preparation, Linux image-building and installation stages. Device preparation or bootloader unlocking can erase user data, and the repository advises backing up the phone and checking manufacturer and application restrictions.
seiso is a Markdown convention and linter for project documentation written by AI and read by people and coding agents. Its convention assigns each document a kind, such as how-to, reference, or ADR; keeps each fact in one canonical location; requires pointers to named files or symbols; separates different concerns; and avoids rapidly changing values in long-lived pages. The reference implementation checks documents against the convention and reports where each violation occurs and how to fix it, including the decision needed when a judgment call is involved. Stable rules run by default, while experimental preview rules require an explicit option; the CLI also provides commands for initialization, rule explanations, parsing, and checks. It can be installed through Cargo, PyPI, npm, or pipx, and integrates with Claude Code hooks, pre-commit, and CI. The project is licensed under MIT.
Splash is a local inference engine for Apple silicon developed by Inco AI. It serves selected language models to coding agents and clients compatible with the OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages APIs, with streaming, tool calls, JSON Schema output, image input, and inline PDF support. Its runtime uses model-specific fused Metal kernels, weights packed for those kernels, a dedicated DFlash 2 draft model for speculative decoding, and a startup-computed memory plan for context, KV cache, and batching. The served model packages include the target model and its draft model; plain MLX or Transformers checkpoints are not supported. Splash runs locally on macOS, binds to a local HTTP server by default, and can be installed through Homebrew. The project is licensed under Apache-2.0, while model weights retain their own licenses.
Three.js Particle Fluids is a GPU particle-physics library for Three.js that simulates and renders liquids, soft bodies, cloth, and smoke with WebGPU. Its API provides a Simulation class, particle rendering, fluid and liquid-surface solvers, soft-body shape matching and mesh skinning, cloth constraints with wind, smoke tracers, colliders, neighbor grids, and mesh-to-distance-field baking. It includes demo presets such as water drops, dam breaks, viscous honey, jelly bodies, cloth, and buoyant smoke, along with documentation and an API reference. It is distributed as the npm package threejs-particle-fluids under the MIT license and requires Three.js r184 and a browser with WebGPU; there is no WebGL fallback.
Tusk is a native, keyboard-driven database client for macOS, Windows, and Linux, written in Rust with the GPUI framework. It connects to databases including PostgreSQL, MySQL, SQLite, SQL Server, Oracle, ClickHouse, Snowflake, BigQuery, Redis, MongoDB, Cassandra, and DynamoDB, and provides a virtualized GPU-rendered data grid, editable tables, schema and DDL tools, SQL editing with language-server completions, query history, exports, backups, SSH tunnels, and connection imports. Its AI assistant reads the relevant table schema, writes a SQL query into a new tab, and explains its assumptions; it does not execute SQL on its own. Tusk can connect the assistant to local agents such as Claude, Codex, or Gemini CLI through the Agent Client Protocol, while its built-in MCP server gives those agents access to an open connection. The application stores credentials in the operating system's credential store and states that it has no account requirement or telemetry. Tusk is free and open source. It is distributed through signed macOS releases and Homebrew, Windows archives, and Linux packages, with source builds available through Cargo. The repository specifies macOS 14 or later on Apple Silicon, Windows 10 or 11, and Linux on a Wayland desktop, with Xwayland supported as a fallback.
Universal Modder is an open-source toolkit by rehan-remade for AI coding agents that create offline, single-player mods for PC games. Its skills and tools identify a game's engine and existing modding route, inspect code and files, back up saves, build a working mod slice, generate art, 3D assets and audio through the fal MCP server, test the result in the running game, package it, and record findings for later agents. The repository supports Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and other agents that read AGENTS.md, and includes the Python `um` CLI for game scanning, asset generation and processing, Windows automation, save management, video capture, publishing checks and a shared knowledge base of game-specific field notes. Its rules restrict use to games owned by the user and offline or single-player work; they prohibit online cheats, anti-cheat or DRM bypasses, and distributing game files or decompiled code. The repository is MIT licensed and requires Python 3.10+ and ffmpeg; Blender is used for 3D-to-sprite rendering.
upm is a small npm-registry package manager written in TypeScript for Node.js. Its command-line operations are also importable JavaScript functions, allowing it to be embedded in other tools. It reads existing .npmrc configuration and supports package-lock.json, pnpm-lock.yaml, bun.lock, and its own upm.lock format. upm resolves and installs dependencies, manages workspaces, runs package scripts and executables, supports registry and tarball dependencies, and shares content-addressed files between projects through a store. It uses hardlinks when available, skips dependency lifecycle scripts by design, supports frozen, offline, production, verification, deduplication, fetching, and pruning modes, and provides an experimental resolver API that can run in a browser without Node.js. The repository describes it as prerelease software with current limitations including no Git or local-directory dependencies and limited peer-dependency handling. It is distributed under the MIT license and is developed in the unjs organization.
Zulu is a native Zulip client for iOS and macOS. It treats topics as first-class: topic-based channels appear as lists of threads, while channels without topics use a plain chat view. Users can organize channels into personal folders, promote topics alongside channels, and rename or hide items locally without changing the Zulip server. The client supports reading and sending messages in channels, topics, and direct messages; replies, reactions, file and image attachments, camera captures, and polls. The repository includes shared SwiftUI code, Zulip API and synchronization packages, a macOS app with sidebar and split-view navigation, and a separate Go notification service that builds to a container but is not deployed. It is described as early but usable and supports sign-in to any Zulip server with password or SSO authentication.
Searchable transcript of GitHub Trending Weekly #51: splash, coucou, dscode, onetake, reelmimic, Rungic, kcc, jeeves, zulu — Github Awesome (14:56). 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 Github Awesome. 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 Welcome back to GitHub awesome. This is GitHub trending weekly number 51. 35 trending open source projects on GitHub right now. Let's go. Running a coding agent on your own Mac sounds great until the model slows everything down. Splash is a local inference engine for Apple silicon that pairs supported models with speculative decoding and custom metal kernels.
00:22 It serves OpenAI and Anthropic compatible APIs, handles vision and tool calls, and connects to agents like Codeex and Open Code. You'll need an M3 or newer Mac and enough memory for your chosen model. Cuckoo puts a little animated character named Mochi in your MacBook notch or your Windows taskbar and has it watch your clawed code sessions. You see each step as it happens, approve or deny permissions without switching windows and drop files straight onto it.
00:53 There's no telemetry and credentials sit in the system keychain. Mochi comes with 28 handcrafted sounds. Dcode is a terminal coding agent for Mac OS that lets its sessions work together. Its shell persists between turns so background jobs keep running while another terminal can inspect or steer the same session. You can hand work to child agents in separate git work trees and have an independent reviewer check the resulting diff.
01:22 The same agent also runs from scripts and CI beyond its terminal interface. Ontake is a claude skill for launch videos shot as one continuous camera move where each beat grows out of the last instead of cutting away. It rebuilds your product's UI in HTML from screenshots so the camera can push into any zoom. Motion blur comes from real multis- sample capture across a 180 degree shutter.
01:48 Every film gets an automatic continuity score before you see it, and anything too cut heavy gets rejected. Real Mimic takes a reference video and breaks down its pacing, framing, colors, and transitions to plan a new 2D animation. You review the storyboard before Claude Code or Codeex agents build shots in parallel with separate agents checking the results.
02:12 Then you can leave notes at exact moments for revisions. It's designed to borrow technique without reusing the references footage or characters. Agent console is a local dashboard for what your clawed code and codec sessions actually cost. It breaks tokens into cache reads, cache writes, output, and uncashed input and shows burn-in tokens per minute and dollars per hour.
02:36 A team can point several machines at one self-hosted hub and only model IDs, token counts, and hashes leave each machine while prompts and file paths stay put. Runic runs a Linux desktop alongside Android on a compatible phone. You can open apps like Firefox, Creda, and Blender, or cast the desktop to a TV. Its bundled assistant can operate those apps on a separate screen while you keep using your phone.
03:04 It's still a private preview tested on a small set of Motorola devices, and preparing a phone can erase its data. Comic pages made for paper can look tiny and washed out on an e-reader. KCC or Kindle comic converter sizes them for your Kindle, Coobo, or Remarkable Screen, trims margins, adjusts black levels, and handles two-page spreads. Feed it image folders, comic archives, or supported PDFs.
03:28 Then export a format your device can read. It's especially handy for DRM free comics you're allowed to convert. Jeves is built for when an app needs a decision. Route a support ticket, flag urgency, or score a customer's frustration. The 9B model reasons first, then returns probabilities for yes or no, multiplechoice, and rating questions through a Jev compatible API.
03:52 Post hog has published the weights, training code, and data. There's a trade-off. Full reasoning takes longer and Jeves trails Jev on some knowledge tests. Tusk is a native database client for Mac OS, Windows and Linux built around a keyboard driven data grid. It connects to Postgress SQL, SQLite, Reddus, MongoDB and more with SSH tunnels and editable rows.
04:17 Ask its AI assistant a question and it reads the schema, writes SQL in a new tab and explains its assumptions. You review the query and decide whether to run it. Hypo lets you test the machinery behind AI scientist workflows before using them for research. It creates synthetic literature with known causal links, then runs mock agents through proposing, debating, ranking, and revising hypothesis.
04:43 Checks flag weak citations and contradictory evidence, while reports show whether the planted links were recovered. Zulu is a native iOS client for Zulip built around the part that can get messy, topics. In channels that use them, Zulu presents each topic as a thread, so conversations stay separate and easier to revisit. In channels without topics, it shows a plain chat instead.
05:08 The app chooses the view automatically, which could make busy Zulip channels easier to follow on a phone. ICode is a terminal workbench for AI agents and workflows that can run fully offline with no telemetry. A workflow shows its progress node by node with each agents model and tool calls in view. And F2 opens any agents instructions, tools, sub aents, and memory for editing.
05:32 You can switch model providers midcon conversation and roll the code back to any earlier turn. Clawed motion design is a skill that builds motion design videos in plain HTML, then captures them frame by frame with playwright and ffmpeg. There's no remotion or gap underneath. Every frame comes from a seek function. Springs are solved in closed form and motion blur comes from blending subframes.
05:57 It builds a beat map from your music and it detects audio peaks to play sound effects. Emagev checks a photo against a record and answers and choices your app can use. Give it a product listing and an image for example and it returns probabilities for which field might be wrong. It can compare two photos, handle texton questions and say when it can't tell.
06:23 The open models run on your hardware, but you'll need to test their confidence thresholds on your own cases. Q Duo turns selected text on your Mac into an action menu beside the cursor. Highlight a sentence in a browser, PDF, or editor. Then translate it, hear it spoken, clean up its formatting, or ask a model about it. The result can replace the selection.
06:45 If the text has moved, Qo copies it instead. Selected text goes to a model only when you choose an AI action. Blueprint before after is a clawed design skill that turns a UI redesign into one continuous animation explaining why each change works. A scan line converts the old screen into a cyan blueprint. Only the changed parts move and resize with guides and dimension lines.
07:09 Then a second sweep reveals the new design. Each fix gets a oneline explanation before the next step. UPM is a tiny TypeScript package manager for npm that you can embed inside another tool. Its commands are importable JavaScript functions so an app can install or resolve dependencies without spawning a separate process. It understands common lock files and your existing npmrc.
07:36 By default, it waits a day before accepting new releases and skips dependency install scripts. Agent office gives your coding agents desks in a shared 3D office. Each GitHub repo becomes a floor, walk up to a worker to watch its live terminal, hand it an issue from the wall, or give it a separate Git work tree. Agents can coordinate through an MCP server, and there's a simpler 2D view for phones.
08:02 It's charmingly literal, though the maintainer warns that features can change between releases. N2E brings hacker news into your terminal, so you can browse without opening another tab. Jump between top, new, ask, show, and the other feeds. Then read comments with Vim keys or a mouse. Collapse long threads, save stories locally for later, and open an article in your browser when it catches your eye.
08:26 It even has a night rider inspired loading animation. 3.js particle fluids makes a browser scene feel surprisingly physical. Water splashes and pours, jelly shapes get squeezed, cloth drapes over objects, and smoke swirls through the air. The simulations run on the GPU through WebGPU with a live demo where you can drop water, change the particle count, and watch rubber ducks float or sink.
08:53 It's a library you can use in your own 3JS scenes. Naive N0.5 Flash is an openweight model built for coding and AI research. It has 309 billion parameters but activates about 15.5 billion at a time and its sparse attention design supports a native million token context window. The weights and inference code are released under MIT. Running it yourself takes serious hardware though.
09:18 The FP8 weights alone occupy roughly 315 GB of GPU memory. Paper Mono shopping list turns an e-aper screen on your fridge into a shared grocery list. Add items from your phone, then take the screen and tick them off aisle by aisle. It works offline and syncs edits when Wi-Fi is available, so a dead zone in the store won't lose your list. There's even an optional claude feature that sorts new items into tackles a quiet problem with AI written project docs.
09:52 A detail changes in one page, but two other pages keep repeating the old version. It gives each fact one home and asks other pages to link there. Its markdown llinter also flags vague references, short-lived numbers in longived docs, and pages mixing instructions with design decisions. Each finding points to the problem and explains what needs fixing.
10:14 OpenAI's MCP extensions add chat GPT specific hooks to MCP so a plugin can feel built into chat GPT instead of bolted on. There are four. a sidebar entry point for your app, file handlers that render your own viewer for a file type, composer mentions that let users search your resources and drop them into a message, and extended forms with things like thumbnail choices.
10:40 Universal modder gives coding agents a playbook for making mods for PC games you own. It identifies the game engine, checks how others have modded it, backs up your saves, then helps build and test a mod in the running game. It can generate sprites, 3D assets, and sound, too. The project is aimed at offline, single player mods, and excludes multiplayer cheats.
11:04 GPUi fast takes the UI framework behind zed and asks a simple question. Why redraw a whole window when only one part changed? This experimental fork keeps unchanged views and layouts from the previous frame, reducing work during scrolling and live updates. It also explores placing a native web view inside a GPUi window while keeping menus and popovers on top.
11:28 Diuramas is a framework for websites where the 3D scene is the page. Its examples let you steer a submarine's lights, wipe condensation off a glass display, or crack open a geode. It pairs an AI image to 3D asset pipeline with a 3JS engine for lighting effects and scroll-driven camera moves. 20 complete example sites give you something concrete to explore and adapt.
11:53 Droid Deck tries to give an Android gaming handheld a Steam Deck feel on supported Adreno devices. It runs Steam in Big Picture mode and uses ARM 64 Proton to launch Windows games with no route required. There's also an optional Linux desktop and a store for ARM 64 Flatpack apps. It's still hardware dependent, so game compatibility and performance will vary from device to device.
12:18 Omari spaces makes the workspace switcher tell you what's actually open. Each workspace shows its app icons. Hover over one for a live preview, then click a window to jump straight to it. If you run clawed code or open code, optional badges show when an agent is working, finished, or waiting for your input. It's a small bar widget that makes a crowded desktop easier to navigate.
12:43 Antigravity fixer is a CLI that diagnoses why you can't get into Google's anti-gravity, the Gemini code assist IDE. It checks the usual culprits, incomplete age verification, stuck phone verification, a stale credential cache, and disconnected OOTH permissions, then clears the expired tokens, and walks you through fresh signin. It never stores your password and only touches local credentials and Google's own pages.
13:10 Job radar watches company career pages so you don't have to keep refreshing them. It spots new openings, marks closed ones, and scores each job against your resume with a short reason. A hard rule pushes senior roles down if you're looking for junior work, even when the skills match. You get daily alerts, a weekly ranked digest, and a local web page to search everything yourself.
13:31 Arc Library is a collection of React components built around subtle physical motion. Buttons respond, charts, transition between data sets, and a billing toggle rolls prices into place. The free repository includes 100 components and 22 larger blocks. You can add one through the Shad CN CLI or copy its source directly, then change the styling yourself.
13:56 Keyboard support and reduced motion behavior are built in. Deck as controller is a deck loader plugin that turns a Steam Deck into a wireless game pad for a Mac or iPad. It pairs over Bluetooth as a PS5 Dual Sense, a PS5 Edge, or an Xbox Elite Series 2. The screen turns off during play. Gyro and the trackpad work, and your Mac shows the deck's battery.
14:20 The back end reconfigures Blue Z to free the HID channels and takes exclusive control of the deck's own gamepad. Cal Kernel is a small language for writing the calculationheavy part of an app. Put typed functions, loops, and number crunching in a CK file, then compile it as a native library or export C or Web Assembly to use elsewhere. Your main app still handles the interface and data storage. It's a focused way to build numerical routines without moving an entire project into a new language.