Agent Skills is an open-source collection of engineering workflows for AI coding agents, developed by Addy Osmani. It packages senior-engineering practices into skills covering the development lifecycle: defining specifications, planning, incremental implementation, testing and debugging, constraints, code review, web-performance auditing, code simplification, and shipping. Slash commands such as /spec, /plan, /build, /test, /review, and /ship activate the corresponding workflows, while skills can also activate automatically based on the work being performed. The /build workflow can generate a plan and implement its tasks in an approved pass, while retaining test-driven verification, individual commits, and pauses for failures or risky steps. It can be installed with the skills CLI or integrated into supported coding agents including Claude Code, Cursor, Codex, Copilot, and Cline.
Code-Graph-RAG is an open-source retrieval-augmented generation system for querying, understanding, and editing multilingual codebases. Its Tree-sitter-based parser extracts functions, classes, methods, modules, and relationships, then stores them in Memgraph under a unified graph schema. An interactive CLI converts natural-language questions into Cypher queries, retrieves source code by name or intent, and supports AI-assisted editing and optimization. It provides AST-based surgical patches with diff previews, structural search and rewriting through ast-grep, dead-code discovery by traversing call and reference edges, and runtime-call overlays from test traces or production eBPF profiles. The project also supports MCP integration and publishes the `cgr` command-line tool to PyPI.
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.
ComfyUI is an open-source AI content-creation engine and diffusion-model GUI, API, and backend developed by Comfy-Org. Its visual node-graph interface lets users build and reuse no-code workflows for image, video, audio, 3D, and text generation, editing, and processing, including workflows using diffusion models, VAEs, text encoders, LoRAs, ControlNets, adapters, and upscalers. Workflows can be saved and loaded as JSON, exposed through App Mode, or integrated into applications through local API endpoints. The engine supports asynchronous queueing, partial graph re-execution, VRAM and RAM management, model offloading, quantized models, custom nodes, and fully offline core operation unless optional resources are requested. It can run locally on Windows, Linux, and macOS through desktop, portable, or manual installations, with support for NVIDIA, AMD, Intel, Apple Silicon, and Ascend hardware. Comfy-Org also provides an official paid cloud version.
DwarfStar is a self-contained, model-specific native inference engine for running DeepSeek V4 Flash, GLM 5.2, and, on high-memory systems, DeepSeek V4 PRO locally. It combines model loading, prompt rendering, tool calls, persistent KV state, an HTTP server, and a coding agent rather than serving as a general GGUF runner. The engine supports Metal on Macs, CUDA including multi-GPU systems, and ROCm on Strix Halo hardware; it also provides SSD streaming for machines with insufficient RAM, distributed tensor and pipeline parallelism, and server-side micro-batching for multi-user inference. The repository includes tooling and data for GGUF, imatrix, quality, and speed testing, and acknowledges code and design contributions from llama.cpp and GGML.
Guava is a set of core Java libraries developed by Google. It adds collection types such as multimaps and multisets, immutable collections, a graph library, and utilities for concurrency, I/O, hashing, primitives, strings, and related tasks. The distribution provides separate JRE and Android flavors. The JRE flavor requires JDK 1.8 or higher; both flavors are published as Maven artifacts under the com.google.guava group ID with the guava artifact ID, and can also be added through Gradle. APIs marked with @Beta may change or be removed, while non-beta APIs are intended to remain binary-compatible.
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.
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.
Ladybird is an independent open-source web browser developed around its own web-standards-based engine rather than an existing browser engine. Its multi-process architecture separates the main user interface, WebContent renderer processes, image decoding, and network requests; each tab has its own sandboxed renderer, while image decoding and network connections run out of process. The project is pre-alpha and intended for developers, with builds for Linux, macOS, Windows through WSL2, and other Unix-like systems. It is licensed under the 2-clause BSD license.
LoopX is an open-source, provider-neutral state kernel and local-first control plane for long-horizon AI agents and peer agent teams. It runs on top of agent harnesses such as Codex, Claude Code, Cursor, dsh, or a custom harness rather than replacing them, keeping objectives, gates, tasks, evidence, quotas, schedules, recovery state, and handoffs durable across bounded turns, sessions, tools, and agents. Its control layer makes semantic decisions about what happens next and supports governance, recovery, human-agent collaboration, and reviewable work. The accompanying personal agent workspace stores goals, attention, conversations, tasks, files, schedules, and recovery state locally across restarts; its supported browser/PWA dashboard can continue work across registered agent sessions and use typed previews, explicit confirmation, and receipts for protected changes.
mise is an open-source command-line tool for managing development tools, environment variables, and project tasks from a single mise.toml configuration file. It installs and switches between tools such as Node.js, Python, CMake, Terraform, Go, and jq, and prepares the configured environment before commands run. Project-specific environment variables can include values from .env files, while tasks define commands and dependencies for activities such as building, testing, linting, validation, and deployment.
npmx.dev is an open-source web frontend for browsing the npm registry, developed as a fast, modern alternative interface rather than a replacement for the registry. It provides package search and details, README and metadata views, source-code browsing with syntax highlighting and line permalinks, generated API documentation when available, version diffs, changelogs, publish timelines, package comparisons, and shareable URL-driven views. It also surfaces dependency and package-health information such as outdated dependencies, OSV vulnerability warnings, provenance indicators, license and install-script warnings, install size including transitive dependencies, download statistics, repository signals, and JSR availability. The interface supports dark and light modes, localization across dozens of locales including right-to-left languages, keyboard workflows, screen readers, and package, team, and organization administration through the local npm CLI.
OmniRoute is an open-source, self-hosted AI gateway developed by diegosouzapw. It exposes model providers through a single OpenAI-compatible endpoint and routes requests according to provider availability, quotas, cost, and latency, with automatic fallback and model chaining. The gateway is designed for local-first operation, keeping provider keys locally and supporting encrypted key storage. Its documented features include quota-aware scheduling, context compression through the RTK and Caveman mechanisms, and integrations for MCP and A2A workflows. The repository describes compatibility with coding clients such as Claude Code, Codex, Cursor, OpenCode, Cline, and Copilot, along with desktop and progressive web app interfaces. It is distributed under the MIT license and supports hundreds of providers and more than a thousand model identifiers.
OpenEdit is an open-source, agent-driven video-editing pipeline from VEED. It has no graphical interface or timeline; an agent-agnostic skill and repository guide drive it through coding agents such as Claude Code, Codex, and Gemini. The pipeline can edit, cut, and reframe footage; create burned-in subtitles, motion graphics, charts, and visual elements; assemble slides, websites, stills, or generated media into video; capture web pages; and connect generation services or MCP servers when needed. Source footage is optional, and VEED Fabric can generate footage only with the user's approval. For captioning, the agent transcribes the video, designs the captions, renders the result, opens a video previewer, and accepts follow-up requests such as changing subtitle color, position, or emphasis. Transcription can use VEED, local offline WhisperX, or a user-supplied service; all providers must produce per-word timings, which are stored in a common transcript format. Caption styles are authored in HTML and CSS. The project includes VEED's closed-source but free-to-use HTML renderer, which does not require a headless browser; Chrome can be used as an alternative rendering backend. V1 primarily targets captions, while motion graphics, charts, and brand-book-matched styling are less exercised. OpenEdit supports Apple Silicon Macs and Windows x64 PCs. Intel Macs are unsupported, Linux support is planned, and Windows requires Git, Node, and ffmpeg. It can be installed from the repository or with `npx skills add veedstudio/open-edit`.
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.
Prime Agent is an open-source command-line coding and research agent for general and long-running tasks, developed by Prime Intellect. It combines recursive language-model workflows through an RLM harness with a persistent Python REPL, treating prompts as variables and invoking tools and recursive subagents programmatically. A continual harness stores supplemental prompts, memories, skill descriptions, and reusable subagent specifications as durable session state that can be refined through small, evidence-backed updates without changing the immutable base system prompt. The agent supports importable Python skills, parallel or background child agents, messaging between running agents, automatic context compaction, persistent goals, heartbeats, schedules, detached sessions, autonomous operation, and retained subagents.
Superpowers is an open-source agentic skills framework and software development methodology for coding agents. It begins by eliciting and presenting a software specification for approval, then produces an implementation plan emphasizing red/green test-driven development, YAGNI, and DRY. After approval, it uses a subagent-driven-development process in which agents implement engineering tasks, inspect and review the results, and continue according to the plan. It is distributed as a plugin for supported coding-agent harnesses, including Claude Code, Codex, Cursor, and others.
Tailwind CSS is an open-source, utility-first CSS framework developed by Tailwind Labs for rapidly building custom user interfaces. Developers compose designs directly in markup using utility classes for layout, spacing, color, typography, responsive behavior, and other styling rules. It supports responsive styling, integrates with PostCSS, and provides documentation at tailwindcss.com along with community discussion through GitHub.
A DataExpert-io GitHub repository that curates learning resources for data engineering. It organizes a 2024 career roadmap, free beginner and intermediate boot-camp materials, required software, hands-on projects, interview guidance, books, communities, newsletters, company and tool directories, engineering blogs, whitepapers, and data-engineering creator accounts. Its categorized references cover areas including orchestration, data lakes and clouds, warehouses, data quality, analytics, data integration, semantic layers, modern OLAP, LLM application libraries, real-time data, and data lineage.
Searchable transcript of Top Dev Tool Projects : Tailwind CSS, ComfyUI, Ladybird, Guava, OmniRoute & mise-en-place — ManuAGI - AutoGPT Tutorials (20:59). 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 DevTool project update video brings them together in one place. Top trending open-source and best DevTool projects this week cover terminal coding agents, utility libraries, local inference engines, browsers, and CSS frameworks. 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 cuttingedge agent frameworks, make sure to check it out.
00:53 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, Prime Agent, self-improving coding agent for long tasks. Prime Agent is an open-source command line coding and research agent built for general and longunning work.
01:15 It runs on a persistent Python environment where file operations, shell commands, tools, and sub aents all happen through code inside a live IPython session. It treats context as variables and spawns real child agents as function calls for parallel or background work. A refined command reviews the current run and saves small evidence-backed lessons as reusable memories, skills, and sub aent specs while never rewriting its base prompt.
01:41 Sessions run in the background through a demon, so goals, schedules, and progress survive after the terminal disconnects. It installs on Mac OS and Linux and connects to subscription or API key model providers. It suits developers and researchers running autonomous multi-step tasks. Install it and run your first session in a project. Project number two, manage agent skills, engineering discipline workflows for AI coding agents.
02:06 Agent Skills is an open-source pack of engineering skills for AI coding agents. It encodes the specs, tests, reviews, and quality gates that senior engineers follow. So, an agent applies the same discipline across every phase of building software. Each skill is a plain markdown workflow with clear steps, verification requirements, and a table of common excuses agents use to skip work, plus rebuttals/ commands mapped to the development life cycle from spec through plan, build, test, review, and ship, and skills activate
02:39 automatically based on the task. You install it through an open skills CLI that supports many agents like Claude Code, Cursor, Codeex, Copilot, and Klein, or as a native plug-in. It suits developers who want their agent to produce production quality code. Add the pack and run your work through the life cycle. Project number three, Superpowers. Full development methodology for coding agents.
03:02 Superpowers is an open-source skills framework and software development methodology for coding agents. Instead of letting an agent jump straight into code, it makes the agent step back, tease out a spec through questions, and show the design in chunks you approve. It then writes a detailed plan and runs a sub aent-driven process where fresh agents work through each task with two-stage review so the agent can work autonomously for long stretches without drifting.
03:27 The skills cover brainstorming, git work trees, testdriven development, systematic debugging and code review, and they trigger automatically as you work. It installs across many harnesses like cloud code, codeex, cursor, gemini cli and open code. It suits developers who want their agent to follow a disciplined process. Install it and let your agent plan before it builds.
03:50 Project number four, Mison Plus one CLI for tools. M tasks. Mison Plus is an open-source command line tool that prepares your development environment before each command runs. It keeps project tools, environment variables, and tasks in one mism file. So, new shells, checkouts, and CI jobs all start from the same setup. It installs and switches between dev tools like Node, Python, Terraform, and Go.
04:18 And it hands back real paths instead of shims. It loads environment variables per project directory, including values from DOP, m files, and it defines and runs tasks for building, testing, linting, and deploying. You install it with a single script, hook it into bash, zsh, fish, or powershell. And it runs on Mac OS, Linux, and Windows. Written in Rust, it draws from a registry of hundreds of tools.
04:41 It suits developers who juggle many languages and projects. Install it and put your whole setup in one file. Project number five, Java. Google's core utility libraries for Java. Java is a set of core Java libraries from Google that fill gaps in the standard library. It adds new collection types like multimap and multi-set immutable collections and a graph library plus utilities for concurrency IO hashing primitives and strings.
05:10 You add it as a Maven or Gradal dependency choosing the JRE flavor for Java 8 and above or the Android flavor for Android and any code that needs to stay compatible with it. The libraries run wherever your Java code runs and are used across most Java projects inside Google and by many other companies. Stable APIs stay binary compatible over time, while ones marked beta may change.
05:34 It suits Java developers who want reliable, well- tested building blocks. Add it to your build and use the utilities in your code. Project number six, DS4, native local inference engine for DeepSeek V4. DS4, also called Dwarfstar, is an open-source native inference engine for running Deepseek V4 Flash and Pro models on your own machine. It is deliberately narrow and self-contained, not a generic model runner.
05:59 And it handles DeepSseek specific loading, prompt rendering, tool calling, and KV state in both RAM and on disk. It runs on metal for Macs, CUDA for Nvidia systems, and RockM for stricks Halo hardware starting from machines with high unified memory. It treats the KV cache as a first class disc citizen. So SSD streaming lets you run models larger than RAM and it can split layers across multiple machines for distributed inference.
06:24 It ships a CLI, an open AI and anthropic compatible server and a native coding agent so tools like codecs, claude code and open code can talk to it locally. It suits developers who want a capable open model running on their own hardware. Build it and run DeepSseek V4 on your machine. Project number seven, the data engineering handbook. Curated learning resources for data engineers.
06:48 The data engineering handbook is an open-source collection of resources for learning data engineering. It gathers in one place a road map for breaking into the field. Free beginner and intermediate boot camp material, hands-on projects, interview advice, and a reading list of more than 25 books. It also links out to tools and companies across orchestration, data lakes, warehouses, quality, OLAP, and streaming, plus blogs, white papers, podcasts, newsletters, glosseries, courses, and certifications.
07:19 Everything lives in plain markdown files you read directly on GitHub with boot camp folders holding notebooks and setup notes. It suits beginners entering the field and working engineers who want a reference of trusted material. Open the repository and start following the road map. Project number eight, Tailwind CSS. Utility first CSS framework for building UIs.
07:41 Tailwind CSS is an open-source utility first CSS framework for rapidly building custom user interfaces. Instead of writing custom stylesheets or fighting predefined components, you compose designs directly in your markup using small single-purpose utility classes for spacing, color, typography, layout, and responsive behavior. It works across the web stack and integrates through post CSS and its own tooling, so it fits into common build setups and frameworks.
08:09 The engine is built in Typescript and Rust, and full guidance lives at the project's documentation site. Because styling stays in your HTML, you can build and adjust interfaces quickly without switching between files. It suits front-end developers and designers who want speed and consistency without leaving their markup. Add it to your project and style your UI with utility classes.
08:31 Project number nine, Omni Route, one endpoint routing across many AI providers. Omni Route is an open-source self-hosted AI gateway that puts hundreds of model providers behind one OpenAI compatible endpoint. You point any OpenAI style tool at a local address, and it routes your request across providers, including many free tiers, so work continues when a quota runs out or a provider fails.
08:55 It works out of the box with no keys by using an auto model that scores connected providers live. And it offers routing strategies for cost, latency, quota headroom, and more. It compresses tool heavy prompts to cut token use. Exposes an MCP server and agent protocol and keeps your keys local. It runs through MPM, Docker, a desktop app, Android via Termox or as a PWA and connects to coding tools like Claude Code, Codeex, Cursor, and Clin.
09:25 It suits developers who want reliable, lowcost access to many models. Install it and route every tool through one endpoint. Project number 10, J-Code terminal coding agent with memory and swarms. J-code is an open-source terminal-based coding agent harness built for multi-session work, deep customization, and speed. Written in Rust, it runs as a fast TUI or as a background server you attach many clients to.
09:51 It gives the agent a semantic memory that embeds each turn, recalls related context automatically, and consolidates itself over time, so you rarely call memory tools by hand. Its swarm mode lets several agents work in one repo at once, sharing messages and resolving file conflicts as code shifts under them. It connects to providers you already pay for through OOTH or any OpenAI compatible endpoint, including local Olama and LM Studio and can resume sessions started in other harnesses.
10:21 A self-dee mode even lets the agent edit, build, and reload its own source. It suits developers running heavy parallel agent workflows. Install it and start a session in your terminal. Project number 11, Code Review Graph, local code map so agents read less. Code review graph is an open-source local first tool that builds a structural map of your codebase.
10:44 So AI coding tools read only what matters. It parses your repository with trees sitter into a graph of functions, classes, and imports linked by calls, inheritance, and test coverage. Stores it in a local SQLite file and serves precise context to your assistant over MCP. When a file changes, it traces the blast radius across callers, dependents, and tests.
11:05 So the agent reviews just those files instead of scanning the whole project. It updates incrementally on saves and commits covers many languages plus Jupyter notebooks and runs as a CLI, an MCP server or a GitHub action that post risk scored pull request reviews. One install command configures many coding tools like Codeex, Claude Code, Cursor and Gemini CLI.
11:28 It suits developers working in large repositories. Install it and build the graph for your project. Project number 12, npmx.dev. Fast modern web browser for npm. npmx.dev is an open-source modern web frontend for browsing the npm registry. It doesn't replace npm. It gives a faster, cleaner interface over the same registry data. You search packages with instant results and open any package to see its readme versions, dependencies, source code with syntax highlighting, generated API docs, version diffs, change logs, and
12:02 a publish timeline. It surfaces health signals like vulnerability warnings, install size, install script and license change alerts, and outdated dependencies. And it can manage your packages, teams, and organizations through your local npm CLI. Swapnpmjs.com in any URL for npmX dev and it just works. And every view is sharable through the URL. Built with NUX, it ships dark mode, keyboard navigation, and a localized accessible interface across many languages.
12:32 It suits developers who evaluate and manage npm packages. Open a package on npmx.dev and explore it. Project number 13. Openedit. Edit videos from your coding agent. Openedit is an open-source video editing skill you drive through your coding agent instead of a rented app. You install it with a single command and it works with claude codeex and Gemini through an agent agnostic skill and repository guide.
12:59 You point your agent at a video and ask it to add subtitles. The agent transcribes the clip, analyzes the transcript, designs a suitable look, and renders the result with captions burned in, then opens a previewer, and tells you where the final MP4 lives. You keep refining in plain language, changing colors, moving text, or emphasizing a specific word.
13:21 Transcription runs through a free V account using your workspace credits, and everything else runs locally with a stored login token. It currently needs an Apple Silicon Mac with other platforms planned. It suits developers who make short videos. Install the skill and ask your agent to caption a clip. Project number 14. Ladybird. Independent web browser with its own engine.
13:43 Ladybird is an open-source truly independent web browser built on a new engine designed around web standards rather than an existing one. It aims to be a complete usable browser for the modern web. It uses a multipprocess architecture with a main UI process, separate renderer processes for web content, and dedicated processes for image decoding and network requests.
14:03 So each tab is sandboxed and decoding, and connections stay isolated from malicious content. Its core is a set of libraries largely inherited from Serenity OS covering web rendering, JavaScript, web assembly, cryptography, and TLS, HTTP, graphics, media playback, and more. Written mainly in C++ with some Rust. It builds and runs on Linux, Mac OS, and Windows through WSL 2.
14:29 It is currently pre-alpha and suited to developers. Build it and help shape an independent browser. Project number 15, Code Graph Rag. Query and edit code bases via knowledge graphs. Code graph rag is an open-source retrieval augmented generation system that turns a multi- language codebase into a knowledge graph you can query and edit in plain English.
14:53 It parses your repository with tree sitter extracting functions, classes, methods, modules, and the relationships between them and stores everything in a mem database under one unified schema. So a monorrepo of mixed languages lives in a single graph. Once built, you can ask questions grounded in the real structure. Retrieve source by name or by intent using semantic search.
15:14 Edit code through ab based surgical patching with a diff preview. Optimize against best practices and find dead code by walking call edges. It translates natural language to graph queries using cloud local or open AI models and runs as an MCP server so tools like cloud code can drive it. It suits developers working in large mixed language repositories.
15:36 Point it at a repo and start asking questions. Project number 16. Comfy UI node graph engine for AI content creation. Comfy UI is an open-source modular engine for AI content creation built around a visual node graph interface. Instead of writing code, you wire together nodes to build workflows that generate images, video, audio, 3D models, and text, and you can reuse them as subgraphs or templates.
16:02 It natively supports many open-source diffusion models and can reach closed models through optional API nodes. It runs locally on Windows, Linux, and Mac OS across Nvidia, AMD, Intel, Apple Silicon, and other hardware through a desktop app, a portable build, or a manual install, and also offers a hosted cloud. It manages VRAMm and RAM smartly, re-executes only the parts of a graph that changed, and supports quantized models, Loras, control nets, and custom nodes.
16:35 Workflows save as JSON, and can even be recovered from generated images. It suits artists and developers building custom generation pipelines. Install it and build your first workflow. Project number 17, LoopX, state kernel for longunning agent teams. Loop X is an open-source lightweight state kernel and local first control plane for keeping longunning AI agents on track.
16:58 Because a model's context is finite, it holds the loop's real state outside the model, the active goal and its scope. Ordered user and agent tudos with ownership. a compact run history with evidence and blockers and quota and scheduling hints. On each turn, it decides whether the agent should act, ask, wait, self-repair, or stay quiet, so work stays reviewable, restartable, and easy to hand off across sessions, models, and hosts.
17:26 It is agent loop agnostic, coordinating codecs, clawed code, and other coding agents through a shared guard without replacing the runtime that does the work. You drive it from a CLI to claim taus log evidence and account for finished slices. It suits developers running agents on tasks that span days. Install it, connect a real project, and start the loop.
17:48 Project number 18, PI agent harness toolkit for building coding agents. PI is an open-source toolkit for building and running AI agents organized as a set of composable packages. At its base is a unified API that talks to many model providers like OpenAI, Anthropic, and Google through one interface. On top of that sits an agent runtime that handles tool calling and state management, an interactive coding agent CLI you run in your terminal and a terminal UI library with differential rendering.
18:19 The coding agent is self-extensible, so it can explain and modify its own behavior. By default, it runs with your own user permissions and no built-in sandbox. So the docs show how to isolate it using a Linux microVM plain Docker or a policy controlled shell. The project treats dependency changes as reviewed code and pins them for supply chain safety.
18:41 Written in Typescript, it installs from npm. It suits developers building or running coding agents. Install the CLI and start a coding session. Project number 19, PI web local browser workspace for PI agent. PI Web is an open-source local web interface for the PI coding agent. It reads the session files PI writes on your machine and gives you a browser workspace instead of scrolling back through terminal history.
19:06 You browse past conversations group by project. Continue from an earlier message or fork a session down a separate path to try a different direction safely. It puts a chat beside a file tree so you can preview source, docs, images, audio, and PDFs while the agent works. and a top bar shows context usage, cost, compaction state, and the system prompt.
19:27 From the web UI, you manage models, login, and API keys, and skills without editing config in the terminal. It reads PI's own session and model files and scopes file access to the selected project. Built-in TypeScript on Nex.js. It runs locally through a single command. It suits developers who use the PI agent. Run it and open your sessions in the browser.
19:49 Project number 20, Kimmyode CLI. Terminal coding agent from Moonshot AAI. Kimmyode CLI is an open-source AI coding agent from Moonshot AAI that runs in your terminal. It reads and edits code, runs shell commands, searches files, fetches web pages, and decides its next step from the results it gets back. It works out of the box with Moonshots Kimmy models, and can be pointed at other compatible providers.
20:14 It ships as a single binary with no node. JS setup starts in milliseconds and runs on Mac OS, Linux, and Windows. You can drop a screen recording or demo clip into the chat and have the agent work from the video. Configure MCP servers conversationally instead of hand editing JSON and install skills, MCP servers, and data sources from a marketplace or any GitHub repo.
20:38 Built-in coder explorer and plan sub aents run in isolated contexts. Life cycle hooks gate risky actions and it plugs into editors like Zed and Jet Brains through the agent client protocol. It suits developers who want an agent in the terminal. Install it and give it your first task. Thanks for watching. See you in the next update.