Agent-evals is a skill designed for evaluating components and end-to-end levels of agentic AI pipeline systems. It enables users to define measurement criteria, build or sample evaluation cases, run repeatable tests, track regressions over time, and derive insights from the results.

Logram is a tool designed to navigate through large logs in the terminal. Users can define the structure of their log lines, filter based on field values, and use advanced search functionalities such as boolean logic and keybindings.

Rudel allows users to collaborate with AI coders like Claude and Codex, providing assistance in writing code, refactoring, and other programming tasks. Users can track their sessions and monitor metrics such as success rate and commit rate.

TablePlus is a GUI tool designed for managing relational databases, supporting MySQL, PostgreSQL, SQLite, and more. It offers features like inline editing, advanced filters, code review, safe mode for production databases, and export/import capabilities. The tool also includes a metrics board for building internal dashboards and a smart query editor with autocomplete and syntax highlighting.

Allows users to build interactive maps using SQL within Snowflake without external tools or data copies. It supports multiple geospatial data types and integrates seamlessly with Snowflake’s security features.

Pictero is a web application that offers a wide range of features for image manipulation, including adjustments like brightness and contrast, annotations, social media export options, OCR capabilities, animation creation, ASCII art generation, and steganography. It supports both manual image loading and integration with several image sources.

Emergent allows users to develop production-ready applications via chat-based conversations, where AI agents handle the design, coding, and deployment process from start to finish.

Skillify is a browser-based tool that reads URLs and generates Skill.md files, which include frontmatter, an overview, patterns, pitfalls, ready to be uploaded as skills for Claude Agent or similar AI tools. It operates client-side without any backend or proxy services.

Pake is a command-line tool that transforms websites into standalone desktop applications for macOS, Windows, and Linux. Built with Rust Tauri, it offers significantly smaller package sizes and faster performance compared to traditional frameworks like Electron.

Provides tools to look up IP address details such as owner information, ASN (Autonomous System Number), geolocation, reverse DNS, CIDR range, and blocklist status. It also allows users to browse Autonomous System Numbers, find organization IPv4 allocations, and view country-specific IPv4 statistics.

Private AI Skills is a service for syncing AI agent skills across different machines, projects, and teams. It provides a single source of truth for skills, ensuring all users always have the latest versions. The service uses end-to-end encryption to keep secrets secure.

This project demonstrates how to integrate Tailscale into a minimal Linux host VM for each container instance on macOS. It runs Tailscale in userspace networking mode, enabling secure SSH connections and MagicDNS without exposing ports or configuring separate SSH servers.

PyFlue allows developers to create agent workflows using Markdown descriptions combined with Python execution. It provides persistent sessions with resumable history, policy-controlled sandbox environments for filesystem and shell access, and validates outputs using Pydantic models. The framework also enables deployment across multiple platforms including Docker, cloud providers, and CI/CD pipelines.

Tyche is a project focused on building a fully autonomous, real-time, distributed, load-balanced, and fair trading bot. It emphasizes optimizing data flows for scalability and handling high-volume trades while exploring the architecture of concurrent systems design.

thr is a small local CLI tool that enables saving explicit memories and recalling them later by meaning or text. It is designed for agent workflows, providing stable JSON output, offline semantic search capabilities, and installable skills for various coding agents.

OpenInbox provides disposable email inboxes via API for programmatic creation, receiving, and verification of emails. It's designed for CI/CD pipelines, QA automation, and staging environments. The service also offers a Chrome extension for easy inbox generation, email field auto-fill, and real-time notifications.

Provides real-time visualizations of music and audio within a web browser. Users can connect their microphone or upload files to see various visual effects based on the sound input.

This tool allows you to manage systemd services through a terminal user interface (TUI). It supports viewing logs, listing services, editing unit files, and controlling the service lifecycle with commands such as start, stop, restart, mask, unmask, enable, and disable. Navigation is similar to Vim.

Julius is a preliminary attempt to create a new dialect of the Lisp programming language implemented using Common Lisp. It requires SBCL and quicklisp dependencies for usage.

Onionlink is a small C++20 Tor v3 onion-service client that communicates directly with Tor relays, builds necessary circuits for v3 access, and supports raw data exchange or HTTP requests with the service. It omits extensive security features present in standard Tor clients to focus on interoperability.

This service allows users to compile and parse custom Magic: The Gathering cards, providing a means to create and test non-official card designs in a digital environment. It includes features such as debugging logs, creating lobbies, and managing game states.

Grinta is a local-first coding agent that plans, executes, validates, and finishes software tasks end-to-end without requiring a cloud control plane or locking users to a specific model vendor. It integrates with local language servers and debuggers, supports multiple LLM providers, and maintains session durability through event-stream ledger. The agent includes safety features such as risk-classified actions, audit logs, and cost/token/latency tracking.

Fabrica is a terminal-based coding agent that executes multi-step tasks using tool calls and provides an interactive TUI with features like scrolling, model switching, and autonomous file editing. It supports multiple providers including Google Gemini, Anthropic Claude, and OpenAI models for code-related tasks.

The service ranks public code repositories on GitHub, GitLab, and Bitbucket based on their compatibility with various AI coding agents. Users can search for repositories by owner or name to see a list ranked according to agent-friendliness scores.

Provides a tool to perform load testing using k6 based on OpenAPI specifications. It allows users to conduct detailed performance analysis for web applications.

The service performs trading analysis using 7 subagents across 5 phases to evaluate stocks by analyzing technical indicators, news, fundamentals, macroeconomic factors, and risk scenarios. It provides structured buy/sell/hold signals with detailed rationale via Claude Code's native Agenttool, using free market data from yfinance without requiring external LLM APIs.

This Rollup plugin provides a more accurate way to bundle TypeScript declaration files (.d.ts) using Microsoft's API Extractor, which is designed not to mangle type names or drop re-exports. It simplifies the process of integrating API Extractor into Rollup workflows by exposing it as an interface.

Provides a solid foundation to start building software as a service (SaaS) products using Flask and Python. Offers the ability to integrate AI features into the product.

Glacier is a modern macOS terminal that rethinks the terminal as a first-class macOS application. It provides auto-tiling panes, file icons, custom background images, live font and theme switching, and a true native feel with familiar copy, paste, and drag functionality. It is built with Rust and Tauri to offer a fast and self-contained experience without requiring system modifications.

Allows users to build their own matchstick puzzles quickly. The platform saves user progress locally on the device, though it may be lost if browser data is cleared or when playing in private/incognito mode.

Provides declarative React components and state helpers for integrating Codex pet spritesheets into web applications. It includes sprite animations, fixed-position rendering with drag-and-drop functionality, and integration with app states via reducers and hooks.

Provides a full-stack framework powered by TanStack Router, featuring server-side rendering (SSR), streaming capabilities, server functions, and more. It supports client-first development while also providing robust server-side features.

This tool converts Docker Compose files into Kubernetes manifests and a UDS Package definition, making them ready for deployment using Zarf. It translates Compose configuration into Kubernetes resources and creates a UDS Package Custom Resource (CR) for network policy, monitoring, and trust-bundle distribution.

Reflect provides a framework that enables AI agents to learn, reason and adapt by observing their runs, capturing feedback, and turning outcomes into reusable skills. It captures every trace an agent produces during its operations, evaluates performance based on set rubrics, and continuously refines strategies for better execution in future tasks.

Nori-Slack-CLI is a command-line interface that allows coding agents to interact with the full Slack Web API. It provides direct access without interactive prompts or ASCII art, returning JSON responses and handling errors efficiently.

Herald is a terminal email client designed to be intuitive and easy to set up, allowing users to read, search, clean up, compose, and manage emails directly from the terminal. It offers context-sensitive shortcut help and optional semantic search and rules for enhanced functionality.

Stanzio is a tool that uses AI to create presentations where each slide is designed using real HTML, allowing for more sophisticated and flexible design options.

Git-issues is a native Git issue tracking tool where issues are stored directly within the repository as plain text Markdown files. This ensures that all issues move with your code and stay version-controlled without relying on external databases or servers.

Responds to triggers like webhooks and cron schedules. Spins up AI coding agents in sandboxed Docker containers with your codebase, reports results back to your chat channel, and can require human approval before actioning. Runs entirely on local machine.

Gnosis creates a searchable knowledge base that records code-related decisions, rejected alternatives, constraints, and intent during AI agent sessions. It integrates with AI agents and stores information as JSONL files with SQLite search indexing, enabling developers to access accumulated context from previous agent interactions.

TermToMD is a tool that transforms messy and inconsistent terminal outputs into structured, readable Markdown. It operates entirely on the client-side, ensuring privacy by not transmitting data.

Provides a catalog of common failure patterns, decision trade-offs, warning signals, and practical playbooks in software development to help engineers avoid typical pitfalls and improve project outcomes.

Cell is a terminal spreadsheet editor that allows users to edit spreadsheets using Vim-like keybindings. It supports basic formula functions and can open, manipulate, and save data in CSV, TSV, or native .cell formats.

Converts inbound emails or cron jobs into structured HTTP POST events. Users can configure endpoints once and receive data in native formats such as Slack, Discord, Teams, Telegram, PagerDuty.

Doccupine converts Markdown or MDX files into modern, visually appealing documentation websites. It supports AI integration and offers customization options such as themes and visual editing features without requiring technical knowledge.

ChatForm allows users to create an AI-powered chat form by first creating a free form on Google Forms and then pasting a generated link. The service converts the Google Form into an AI chat form. It emphasizes ease of use without requiring user accounts, tracking, or payment.

MemOperator-4B is a language model designed specifically for managing memory within the MemOS system. It excels in tasks such as extracting, organizing, and updating memories from both conversations and documents. The model offers faster performance with lower resource consumption compared to larger models like Qwen3.

Detects hidden characters in code that can be used to hide executable payloads, flags sequences of invisible characters as potential threats, and identifies stray invisible characters from copy-paste operations. Runs entirely client-side with no data leaving the user's machine.

Demonstrates how fake diff text within commit messages can be applied as real patches when using GNU patch with GitHub's .patch URLs. Shows that commit message content can inadvertently create files during patch application.

A dead-simple SQLite-backed job queue that enables cross-device script coordination with no need for complex infrastructure like Redis or RabbitMQ. It allows any script to post jobs, which can be picked up and executed by other scripts on different machines or devices.

Aether is an open-source core compiler for the Aether Agent Reliability Framework. It replaces fuzzy prompts with strict Weighted Intent Token (WIT) vectors to prevent Context Rot. The framework helps maintain consistent behavior in LLM agents by using precise vector representations instead of ambiguous natural language prompts.

CrabPDF is a free, browser-based tool for editing PDF text, running OCR on scanned documents, finding and replacing text, splitting poorly detected text fragments, and redacting documents without uploading files to a server.

Polynya allows users to stream data from PostgreSQL databases to Iceberg format, creating isolated analytical environments for AI agents. It provides a fresh and secure querying layer without impacting production data.

Auge Vision allows users to analyze images through Apple's Vision framework using a UNIX CLI in their terminal. It provides OCR text extraction, classification labels, barcode decoding, and face bounding boxes with a single command.

Provides a framework-agnostic solution to create Keynote-style text trace animations using SVG. It offers bindings for Vue and React, along with features like per-glyph styling, accessibility support, and integration options for custom fonts.

<hyper-frame> is a custom element that acts as an iframe capable of embedding any website. It is powered by a live remote browser, providing features such as an embeddable surface, navigation controls, automation capabilities, screenshots, tab events, and streaming health metrics. This allows developers to embed and interact with websites that typically cannot be framed due to security restrictions.

DSS is a command-line tool that allows for editing spreadsheets in a full-screen terminal interface. It supports loading DSS, CSV, XLS, XLSX, and XLSM formats and offers features such as formula computation, real-time cell editing, and exporting back to DSS.

Axle is a continuous integration (CI) tool that automatically checks pull requests on GitHub for WCAG 2.1 / 2.2 AA violations and suggests real code fixes using Claude AI. It generates legal artifacts required by regulations such as the EAA 2025, ADA, and Israeli equality laws.

Implit is a tool that scans AI-generated code and validates every import before it's run to prevent broken builds caused by non-existent or incorrectly imported packages.

Provides a Rust-based implementation of the Hyperswarm peer-to-peer networking protocol that is fully interoperable with existing Node.js networks. Includes Kademlia DHT, Noise handshakes, and reliable UDP transport.

Tabby is a Chrome extension that helps manage an overwhelming number of browser tabs. It offers features like duplicate detection by full URL, the ability to close redundant tabs while keeping one per group, exporting unique tab links, and organizing tabs into smart categories with preview and safe limits.

LoreData creates personas based on characters from popular fictional worlds like Breaking Bad, Game of Thrones, and more. It provides deterministic output without network requests, allowing users to generate consistent data in Node.js, browser environments, and as a CLI tool.

MCP Spine is a local-first proxy that sits between an LLM and MCP servers, providing security, routing, token control, and compliance. It offers features like rate limiting, secret scrubbing, semantic routing, schema minification for token savings, state guard for file version control, and a plugin system for custom middleware hooks.

Provides a retro-themed React component that displays an old-school modem terminal and synthesizes dial-up handshake sounds using the Web Audio API.

MockHop allows users to capture any webpage using a built-in browser without needing extensions or additional apps, then annotate the captured images with various tools like arrows, rectangles, highlights, and text callouts. Once annotated, users can export high-resolution PNG files for use in Slack threads, design reviews, or bug reports.

Aethergent provides a simple and immediate way to generate Ethereum wallets for AI agents through a single HTTP POST request, offering an alternative to traditional wallet-as-a-service providers which require multi-step onboarding. It returns an Ethereum address and private key instantly without any authentication or setup.

Mdspec allows teams to automatically publish and manage Markdown files in various destinations such as ClickUp, S3, Notion, Confluence, and more. It requires a mapping file placed within the repository and integration steps through CI tools like GitHub Actions.

Bulk URL Checker allows users to upload a CSV file containing up to 75,000 URLs. It checks each URL for status codes, redirects, and response times in the cloud, sending an email report when completed.

This assembly-based tokenizer parses more than 1 gigabyte of ASCII per second by removing whitespace and separating strings with a null terminator. It is designed for SSE2 CPU architectures and can process text at speeds up to 972MB/s.

SVG Fitter is a vector image processing service that converts images into SVG format. It provides controls for adjusting tolerance, simplifying shapes, and segment settings to customize the vectorization process. The service also offers options for shape smoothing and filling holes in vector images.

Mux0 is an open-source terminal application designed specifically for macOS. It features workspaces, tabs, and split panes to manage repositories efficiently. Additionally, it provides real-time status updates for development tools like Claude Code, OpenCode, and Codex.

뷰 전환 기능을 쉽게 사용할 수 있도록 다양한 유틸리티 함수를 제공합니다. 이 툴킷은 피처 감지, 시뮬레이션 지원, 애니메이션 유틸리티, 전환 재생 제어 등 여러 모듈로 구성되어 있습니다.

SpaceSim95 provides a lightweight platform for testing orbital geometry, coverage analysis, and network design scenarios. It streamlines workflows by eliminating nested menus and allowing iterative design approaches in early mission planning.

Provides a text-based user interface (TUI) to manage and interact with the local development environment in Tilt, allowing real-time status updates, resource management, and detailed views without leaving the terminal.

NoonFlow is a macOS application designed to enhance developer productivity by integrating AI tools like Claude Code and Codex. It provides features such as server management, dashboard optimization, and improved interaction with external services.

Provides a command line tool that helps users organize and manage their aliases for frequently used commands. It supports installation via Homebrew, manual download, and adding the script to startup files for persistent usage.

Bunny Agent is a coding agent built on Pi Coding Agent, enabling developers to build and deploy coding agent SaaS solutions via a native AI SDK UI. It supports multi-model AI runners (Claude, Gemini, OpenAI, etc.) and integrates with remote sandboxes for development tasks.

Stash provides a memory layer for AI agents that allows them to remember user interactions across sessions using PostgreSQL and pgvector. It helps maintain continuity in conversations without requiring users to repeat information.

Useknockout provides a background removal API that processes images to produce transparent PNG outputs. It uses BiRefNet, a state-of-the-art model for background removal tasks, to deliver high-quality results with fast processing times and at a significantly lower cost than competing services like remove.bg.

LLMCat is a command line interface tool designed to automatically format and clean up messy codebases, making them suitable for use with large language models (LLMs). It supports multi-file processing and allows users to configure output preferences through settings files.

Codex is a tool that significantly reduces the size of software engineering benchmark (SWE-bench) traces. It aims to minimize 'context bloat' in these traces, making them more efficient and easier to analyze.

Lightwhale is an operating system specifically tailored to run Docker containers efficiently. It live-boots from an ISO directly into a fully operational Docker Engine, thereby eliminating the need for installation or configuration. The core system remains immutable, which enhances security and minimizes maintenance.

HNswered is a Chrome side panel extension that monitors public Hacker News stories and comments you have authored. It detects replies, even when the original comment is deep in a thread, and displays them in a clean inbox with unread/read filters. The application maintains its state locally within Chrome storage without requiring any server interaction or login credentials.

Offers a service for parallel video rendering and provides users with a live dashboard to monitor the progress.

cctape is a local proxy that archives every request and response made through Claude Code, offering features such as full-text search across sessions, tracking of token usage and cost, and an MCP server for Claude to access past session data.

The service converts graphical designs, such as screenshots or sketches of web pages, directly into HTML markup. Users can optionally sketch parts of their design to guide the conversion process.

Provides a full-featured database workspace that can be run on users' own servers. It utilizes Docker for setup and ensures data remains within the user's network.

Figpack is a Python library that enables users to generate interactive and shareable HTML visualizations from Python scripts. It supports local creation of visualizations without authentication but requires an API key to upload figures to the cloud for sharing via URL.

Stvor provides Signal Protocol encryption (X3DH + Double Ratchet) for applications, offering a dead-simple API that requires no cryptography expertise. It works in both Node.js and the browser environments.

Verifies that the SDocs website serves the same code available in its GitHub repository, ensuring client-side privacy and security for sensitive data handled within Markdown files. This process uses SHA-256 hashing to compare file contents.

DeepSeek V4 is an advanced AI model designed to support extensive context with a large memory capacity of up to 1 million words. It offers quick response times, enhancing user interaction.

MirrorNeuron is a platform that allows users to create and run reliable AI workflows using pre-built blueprints, simplifying the process from initial setup to deployment on various platforms like laptops, edge nodes, clusters or cloud environments.

GeoValida uses satellite data and analysis layers such as livability score, vegetation change, surface temperature, and investment zones to provide a ranked list of the best available land based on criteria like greenery, nearby services, road access, and growth momentum. Users can draw an area on the map to receive detailed insights.

Pdfnative is a TypeScript PDF engine that does not require any dependencies. It appears to be designed to work within the npm ecosystem, as it is hosted on www.npmjs.com.

Endo Familiar is a working prototype of an object-capability based JavaScript runtime for AI agents. It provides a secure environment where agents can operate with controlled permissions, running in sandboxed compartments with carefully scoped access. The framework ensures agents only have the minimum authority needed to perform their specific tasks.

Tolaria is a desktop app for Mac that helps users manage markdown knowledge bases. It allows users to organize notes as plain markdown files, which are portable and can be edited with any text editor. The app supports version control through Git, enabling full version history and use of any Git remote without dependency on Tolaria servers.

Noemica designs representative personas, runs them through production, staging, or local sites, and provides feedback on why they discontinued use in plain language. It allows for early detection of user drop-offs without the need to wait for real user feedback.

PViz is used to trace, measure and refactor large-scale dependency cycles within complex Python projects like Scrapy. It helps identify critical modules that form an architectural kernel, enabling systematic removal of redundant dependencies without breaking core functionalities.

Provides elegant default styling for all standard HTML elements without requiring additional classes. Includes dark mode and integrates well with existing styles.

Hodor is a Windows 11 credential provider DLL that accepts unlock commands over a named pipe. Any application that can write to a pipe can unlock the lock screen or approve a credential prompt without requiring Windows Hello enrollment or built-in biometric hardware. The DLL runs inside LogonUI.exe on the lock screen and hands credentials to Windows for validation.