AI Agents & Automation

adk-js is an open-source, code-first TypeScript toolkit designed for developers to build, evaluate, and deploy sophisticated AI agents with flexibility and control. It allows for defining agent behavior, orchestration, and tool use directly in code, facilitating robust debugging, versioning, and deployment across various environments. Key features include a rich tool ecosystem for integrating pre-built tools, custom functions, and OpenAPI specs, as well as support for modular multi-agent systems. The toolkit is particularly useful for developers seeking fine-grained control and tight integration with Google Cloud services, offering a built-in development UI for testing and debugging agents.

Agently is an open-source GenAI application development framework designed to build production-grade AI applications. It offers stable outputs through contract-first schema enforcement and automatic retries, testable orchestration with TriggerFlow, and observable actions via comprehensive logging of tool and sandbox calls. The framework supports structured streaming for instant event processing and provides a flexible Action Runtime for functions, MCP servers, and sandboxes. Agently's TriggerFlow enables serious workflow orchestration with concurrency, event-driven branching, human-in-the-loop interrupts, and execution persistence. It also includes robust session management for multi-turn memory and project-scale configuration management with hierarchical settings.

Amazon Bedrock AgentCore Samples provides a comprehensive collection of examples and tutorials to help developers understand, implement, and integrate Amazon Bedrock AgentCore capabilities into their applications. This tool is designed to accelerate AI agents into production by offering the scale, reliability, and security critical for real-world deployment. It is both framework-agnostic (supporting Strands Agents, CrewAI, LangGraph, LlamaIndex, etc.) and model-agnostic, allowing users to bring their preferred Large Language Models. Key features include a secure, serverless runtime for deploying agents, a gateway to convert APIs and Lambda functions into MCP-compatible tools, managed memory infrastructure, and built-in tools like Code Interpreter and Browser Tool. The repository also offers resources for deployment automation, observability, evaluation, and fine-grained access control with Cedar policies.

Artus is an AI Product Manager designed to help product leaders build applications based on informed, data-driven decisions. It ensures that AI-driven coding aligns with the user's vision by leveraging specialized AI agents for business analysis, UI/UX design, and full-stack engineering. These agents evaluate every decision across business, design, and engineering, identifying gaps, conflicts, and risks before execution. Artus provides product foresight and oversight, offering pre-execution risk and impact reports, end-to-end decision traceability, and engineering-ready architectures. It is built for individuals and teams who prioritize alignment, feasibility, and long-term impact, allowing for granular control over product decisions, deep research capabilities, design calibration, and impact analysis.

Context Space offers a comprehensive context engineering infrastructure designed to enhance the productivity of AI agents, automation workflows, and developer tools. It provides unified MCP (Model Context Protocol) tools and secure, verified integrations, allowing AI agents to access real-world data and services effectively. The platform addresses challenges like manual credential handling with one-click OAuth and vault security, inconsistent APIs with a single RESTful API, and complex deployments through a unified context plane. Context Space is evolving into a complete context engineering platform, supporting advanced AI capabilities like semantic retrieval, context optimization, and real-time updates, making AI agents truly usable and secure.

Kompas AI is a personal knowledge management tool designed to help users capture, organize, and leverage their thoughts and research. It integrates AI sessions that use past notes and reading to ask better questions, fostering a deeper engagement with content. The tool builds a graph of your thinking, where every session becomes nodes and edges, connecting decisions, memories, and retrieved passages. This graph then aids in retrieval-first writing, allowing users to pull sessions and fragments into their drafts, ensuring their unique voice and context are preserved. Kompas emphasizes data ownership, with content stored locally in a SQLite file, and offers both a free desktop app and a forthcoming Pro cloud tier with included AI.

Wedge serves as an operating system for healthcare AI agents, specifically designed to automate the entire back office for large health systems. It offers tailored AI solutions, handling development, integration, and maintenance. The process begins with a comprehensive audit to identify automation opportunities, followed by defining clear ROI for each AI agent. Onsite engineers then customize and develop these agents, moving department by department to automate tasks. Key offerings include AI Payment Reconciler, AI Records Retrieval, AI Governance, AI Receptionist, AI Claims Manager, and AI Medical Coder, all aimed at improving efficiency, reducing costs, and maximizing revenue recovery for healthcare organizations.

UNOY is an AI-assisted workspace designed for organizations working with regulated knowledge, such as law firms, chambers, associations, and back-office teams. The platform allows users to transform their expertise into digital guides and automated workflows, eliminating scattered knowledge in emails and PDFs. COIA, the central AI orchestrator, coordinates specialized Work Agents to process incoming requests, gather data, verify information, and prepare drafts or reports. This system ensures tasks are completed with defined processes, deadlines, and an audit trail, allowing human teams to review and approve results. UNOY aims to scale knowledge, automate back-office operations, and provide a structured approach to work.

DeerFlow is an open-source SuperAgent harness designed to orchestrate sub-agents, memory, and sandboxes for complex tasks like research, coding, and creation. It leverages extensible skills to handle a wide range of tasks, from those taking minutes to those requiring hours. The platform supports various LLM providers, including OpenAI, OpenRouter, and vLLM, and offers flexible deployment options via Docker or local development. Key features include a robust sandbox and file system, long-term memory, and context engineering, making it suitable for developers and researchers building advanced AI applications. DeerFlow 2.0 is a ground-up rewrite, focusing on enhanced capabilities and a streamlined architecture.

ToothFairyAI offers a private and secure AI Studio where autonomous AI agents can operate under your control, ensuring data sovereignty and privacy. These agents are designed to enhance productivity and streamline operations by planning, reasoning, and performing tasks 24/7. The platform supports multimodal agents capable of creating and analyzing images, videos, audio, and even rendering 3D models. ToothFairyAI agents integrate seamlessly with existing systems like SharePoint, Google Workspace, Microsoft 365, CRM, and project management tools, leveraging advanced capabilities such as code execution, long-term memory, and API connections. It provides specialized agents for various business needs, including customer service, data analysis, project orchestration, and sales, with a focus on zero data retention and transparent data handling.

docling-api is a robust and scalable backend server designed for efficiently converting a wide range of document formats into Markdown. Leveraging Docling (IBM's advanced document parser) and built with FastAPI, Celery, and Redis, it supports formats including PDF, DOCX, PPTX, HTML, images (JPG, PNG, TIFF, BMP), AsciiDoc, and CSV. The server provides features like text extraction, table detection and conversion, image extraction, and multi-language OCR. It offers both synchronous and asynchronous API endpoints for single and batch document conversions, with job tracking for asynchronous tasks. Optimized for both CPU and GPU processing, with GPU recommended for production, docling-api is ideal for large-scale workflows requiring high performance and flexibility in document processing.

ebook-GPT-translator is a modernized, open-source toolkit designed for translating ebooks and documents across various formats including TXT, EPUB, DOCX, and PDF, with optional MOBI support. It leverages several AI translation providers such as local Codex CLI, Claude Code, Gemini CLI, and OpenAI/Azure APIs, along with compatible OpenAI endpoints. Key features include a resume-safe SQLite translation cache, configurable chunk and token limits for long books, and support for glossary files (CSV/XLSX) to ensure terminology consistency. The tool also offers chapter memory, rolling translated context, and automatic term memory to improve long-novel consistency, making it ideal for users who need high-quality, style-preserving translations.

Augen Pro is at the forefront of engineering AI-driven, general-purpose computing designed to enhance everyday life. The company focuses on "Pro-Human" invisible computing, aiming to make technology work effortlessly for users and unlock limitless potential. Their research and development is currently centered on two innovative devices: A¹ Sense and B¹ Eye, which are part of their wearable and neural technology initiatives. Augen Pro's mission is to create smarter, smaller AI health tools and lead the future of Invisible Computing, simplifying Heads-Up Computing industries. They prioritize safety, accessibility, and reliability, ensuring their advanced technological devices are seamlessly integrated with the human body.

Eland is a Python client and toolkit designed for seamless interaction with Elasticsearch, offering a Pandas-compatible API for data exploration and analysis. It enables users to work with large datasets residing in Elasticsearch without loading them into local memory, making it ideal for big data applications. Beyond data manipulation, Eland facilitates machine learning workflows by allowing users to upload trained regression and classification models from libraries like scikit-learn, XGBoost, and LightGBM directly into Elasticsearch for inference. It also supports importing PyTorch-trained BERT models for NLP tasks, including those from the Hugging Face model hub, enhancing Elasticsearch's capabilities for advanced AI applications and ETL processes.

Fulling is an AI-powered full-stack engineer agent designed to streamline the development process. It offers a sandboxed environment complete with Claude Code as an AI pair programmer and a dedicated PostgreSQL database. The platform automates the setup of necessary infrastructure, allowing developers to focus on coding. Key features include a full-stack development environment, a web terminal, a file manager, and live HTTPS domains. It supports configuration-driven development, enabling users to integrate services like Stripe or OAuth by simply entering API keys, with Claude Code implementing the features. Fulling also provides GitHub integration for version control and one-click deployment from sandbox to production.

Macaron AI is positioned as the world's first personal AI agent, designed to enrich your lifestyle rather than just boost productivity. It functions by creating personalized tools and mini-apps for various aspects of life, including travel, health, relationships, and hobbies. A key differentiator is its "Deep Memory" feature, which allows Macaron to grow with the user and remember important details, much like a real friend. The platform emphasizes providing real-life solutions through simple requests, eliminating the need for frustrating adjustments. It aims to build tools users actually need, making it a unique personal assistant that focuses on improving daily living.

google-search is a Playwright-based Node.js tool designed to perform Google searches and extract results while bypassing common anti-scraping mechanisms. It offers a local alternative to paid SERP APIs, providing advanced techniques like intelligent browser fingerprint management, automatic browser state saving, and smart headless/headed mode switching to reduce detection risk. The tool can function as a command-line utility for direct searches or as a Model Context Protocol (MCP) server, enabling AI assistants like Claude to access real-time search capabilities without needing additional API keys. It supports raw HTML retrieval, page screenshots, and outputs results in JSON format, making it highly customizable and extensible for developers.

FlagGems is a high-performance, generic operator library implemented in the Triton language, designed to accelerate Large Language Model (LLM) training and inference. As part of FlagOS, it aims to unify model-system-chip layers, enabling a "develop once, run anywhere" workflow across various AI accelerators. This approach unlocks hardware performance, eliminates fragmentation among AI chipset-specific software stacks, and significantly lowers the cost of porting and maintaining AI workloads. FlagGems offers a large collection of PyTorch compatible operators, hand-optimized performance for selective operators, and eager-mode readiness. It supports automatic pointwise operator codegen, fast per-function runtime kernel dispatching, and a multi-backend interface for diverse hardware platforms, including over 10 supported backends.

World Simulator AI offers an engaging platform for users to dive into immersive, AI-powered virtual worlds. Players experience stories in first person, with their choices directly influencing the narrative's progression. The tool supports a wide array of genres, from historical conquests like Alexander's Conquest and Pharaoh's World to fantasy adventures such as The Last Sorceress and Beast-Tamer’s Trial, and even horror scenarios like Teddy Bear Chase. Users can explore pre-made worlds or create their own, offering endless possibilities for interactive storytelling and roleplay. This platform is ideal for those who enjoy 'choose your own adventure' style narratives and want to experience dynamic, AI-driven stories.

gpt-oss is a series of open-weight language models developed by OpenAI, designed for advanced reasoning, agentic tasks, and diverse developer use cases. It includes two primary models: gpt-oss-120b, suitable for production and general-purpose high-reasoning tasks on a single 80GB GPU, and gpt-oss-20b, optimized for lower latency and specialized local applications within 16GB of memory. Both models are trained with a harmony response format, which is crucial for their correct operation. Key features include a permissive Apache 2.0 license, configurable reasoning effort, full chain-of-thought access for debugging, fine-tunability, and agentic capabilities like function calling and web browsing. The models also utilize MXFP4 quantization for efficient memory usage.

LaunchLemonade offers AI agents designed to automate back-office operations for regulated professional services firms, including accountants, financial advisors, consultants, and fractional CFOs. The platform provides pre-built agents and workflows for tasks like meeting intelligence (notes, action items, follow-ups), email management (triaging, scheduling), and searchable knowledge bases from firm-wide data. It emphasizes compliance and governance, with features like audit logs, role-based access, and data retention controls built-in. LaunchLemonade also offers services for custom agent development, AI consulting, team training, and white-labeling the platform, enabling firms to integrate AI safely and efficiently without needing an in-house engineering team.

SageFlow is a no-code platform designed for building and deploying powerful AI agents using a drag-and-drop interface. Users can create custom agents with pre-built blocks, enabling rapid development and deployment. The platform also features an AI Marketplace where users can discover and rent expert-crafted AI agents for various tasks, or monetize their own creations. Key benefits include predictable income for creators, effortless scaling of agents, and flexible payout options. SageFlow aims to simplify agentic AI for everyone, offering solutions for tasks like SEO optimization and social media strategy, with future integrations planned for voice assistants, mobile devices, and a community platform.

Human Emulator, powered by Grok, offers AI digital workers capable of performing any computer task a human can do, but faster, cheaper, and at infinite scale. This includes data entry, customer emails, invoice processing, and report generation. The platform automates back-office tasks, handles finance and compliance operations, supports software development with AI teams, provides business analytics, generates content and documents, and manages customer operations. Users describe the work needed in plain English, receive an instant quote based on complexity and volume, and can deploy and scale workers instantly, handling 10x volume with zero ramp-up.

LongLLaMA is a large language model specifically designed to manage and process exceptionally long contexts, up to 256k tokens or more. Built upon the OpenLLaMA foundation and enhanced with the innovative Focused Transformer (FoT) method, it allows language models to handle extensive inputs while training on shorter sequences. The FoT method uses contrastive learning to enable attention layers to access a memory cache, significantly extending the effective context length. LongLLaMA is available in several variants, including a 3B base model under an Apache 2.0 license, and instruction-tuned versions like LongLLaMA-Instruct-3Bv1.1. A LongLLaMA Code 7B model, based on Code Llama, is also provided for code-related tasks. The project offers inference code, instruction tuning, and FoT continued pretraining code, making it a valuable resource for researchers and developers working with large language models and context scaling.