1. 666ghj/MiroFish Not new today
A Simple and Universal Swarm Intelligence Engine, Predicting Anything. 简洁通用的群体智能引擎，预测万物
First seen: March 08, 2026 | Streak: 2d
Analysis
MiroFish is a swarm intelligence engine designed to simulate and predict complex real-world scenarios by creating high-fidelity digital environments populated by autonomous agents. Technically, the platform leverages GraphRAG for memory construction, entity relationship extraction, and multi-agent simulation to analyze input seeds such as news, policy drafts, or literature. By dynamically injecting variables into these "digital sandboxes," the system allows users to observe social evolution and emergent behaviors to generate detailed predictive reports and interactive simulations.
This project is primarily aimed at decision-makers requiring low-risk policy pre-simulation, as well as creative users looking to explore narrative outcomes or hypothetical scenarios. It is gaining traction because it bridges the gap between abstract data analysis and concrete, visual interaction, turning static information into a dynamic "God’s-eye view" of potential futures. By democratizing access to complex multi-agent simulations, MiroFish offers a versatile, "playful" yet powerful tool for both professional strategic planning and personal intellectual curiosity.
2. thedotmack/claude-mem Not new today
A Claude Code plugin that automatically captures everything Claude does during your coding sessions, compresses it with AI (using Claude's agent-sdk), and injects relevant context back into future sessions.
First seen: February 02, 2026 | Streak: 1d
Analysis
Claude-Mem is a sophisticated plugin for Claude Code designed to provide persistent, long-term memory for coding sessions by capturing and summarizing tool interactions and project context. It operates through a robust architecture featuring lifecycle hooks, a Bun-managed worker service, and a hybrid search engine that combines SQLite for relational data with Chroma for semantic vector search. By utilizing a "progressive disclosure" retrieval pattern, the system intelligently fetches context in layers to minimize token consumption while maintaining deep project continuity across disconnected sessions.
This tool is highly beneficial for professional software engineers and heavy users of Claude Code who struggle with context loss during long-term development projects. It is trending because it addresses a critical pain point in AI-assisted coding—the ephemeral nature of LLM sessions—by enabling the agent to "remember" past bugs, architectural decisions, and previous code iterations. The combination of automated background operation, privacy controls, and a dedicated web UI makes it an essential utility for developers looking to scale their AI-driven productivity.
3. Crosstalk-Solutions/project-nomad Not new today
Project N.O.M.A.D, is a self-contained, offline survival computer packed with critical tools, knowledge, and AI to keep you informed and empowered—anytime, anywhere.
First seen: March 16, 2026 | Streak: 2d
Analysis
Project N.O.M.A.D. is a self-contained, offline-first server designed to provide users with critical tools, knowledge, and AI capabilities without requiring an active internet connection. It functions as a management UI that orchestrates various containerized applications—including Kiwix for archives, Kolibri for education, and Ollama for local AI—using Docker on Debian-based operating systems. By streamlining the installation and configuration of these services, it effectively transforms hardware into a robust, portable "command center" for information and utility.
This project is ideal for preppers, educators in remote regions, researchers, or anyone seeking to build a digital contingency plan that remains functional during infrastructure outages. It is gaining traction because it democratizes access to sophisticated AI and vast educational datasets, offering a practical solution for maintaining information sovereignty in an increasingly digital world. As the reliance on cloud services grows, N.O.M.A.D. provides a compelling, privacy-focused alternative that empowers individuals to maintain control over their data and resources regardless of external connectivity.
4. obra/superpowers Not new today
An agentic skills framework & software development methodology that works.
First seen: February 04, 2026 | Streak: 7d
Analysis
Superpowers is an agentic framework designed to standardize the software development lifecycle by integrating a library of composable, automated "skills" directly into coding agents like Claude Code and Cursor. Instead of jumping straight to code, the system enforces a structured workflow that begins with Socratic brainstorming, proceeds to detailed implementation planning, and culminates in a rigorous subagent-driven development cycle. Technically, it functions as a plugin that triggers mandatory engineering protocols—such as strict test-driven development (TDD), git worktree management, and iterative code reviews—ensuring that agents maintain high-quality standards and consistent progress without requiring manual intervention.
This project is highly beneficial for developers and software teams looking to minimize the "hallucination" and ad-hoc coding patterns often associated with AI-assisted programming. It is trending because it addresses a critical gap in the current AI developer tool ecosystem: the transition from simple code generation to reliable, autonomous project execution that mirrors professional engineering discipline. By embedding best practices like YAGNI and DRY into the agent's core decision-making loop, Superpowers provides a repeatable, scalable methodology for building robust software applications.
5. abhigyanpatwari/GitNexus Not new today
GitNexus: The Zero-Server Code Intelligence Engine - GitNexus is a client-side knowledge graph creator that runs entirely in your browser. Drop in a GitHub repo or ZIP file, and get an interactive knowledge graph wit a built in Graph RAG Agent. Perfect for code exploration
First seen: February 22, 2026 | Streak: 2d
Analysis
GitNexus is a zero-server code intelligence engine that transforms local codebases into interactive knowledge graphs for enhanced AI developer assistance. By leveraging Tree-sitter for code parsing and a persistent, high-performance database called LadybugDB, it indexes dependencies, call chains, and execution flows directly on the user's machine. The tool operates as an MCP (Model Context Protocol) server, providing AI agents—such as Claude Code and Cursor—with deep architectural awareness, impact analysis, and automated refactoring skills without requiring external data transmission.
This project is primarily designed for professional software engineers and AI-assisted development teams who struggle with the context limitations of standard LLMs. It is currently trending because it solves the "blind edit" problem, allowing agents to understand the broader implications of code changes and dependencies that otherwise result in broken features. By offering both a visual web-based explorer and a robust CLI for seamless editor integration, GitNexus significantly bridges the gap between static code analysis and reliable, agent-driven software development.
6. lightpanda-io/browser Not new today
Lightpanda: the headless browser designed for AI and automation
First seen: March 14, 2026 | Streak: 4d
Analysis
Lightpanda is an open-source, headless browser specifically engineered for high-performance automation, AI agents, and web scraping tasks. Built using Zig and the V8 JavaScript engine, it provides a lightweight alternative to traditional browsers by offering ultra-low memory usage, instant startup, and significantly faster execution speeds. It supports standard Web APIs and exposes a Chrome DevTools Protocol (CDP) server, allowing it to integrate seamlessly with existing automation frameworks like Puppeteer, Playwright, and chromedp.
Developers and AI engineers focused on large-scale web interactions will benefit most from Lightpanda’s efficiency when running hundreds of concurrent browser sessions. The project is gaining traction because it solves the resource-intensive overhead typically associated with headless Chrome, enabling more cost-effective and scalable automation on cloud infrastructure. By providing a streamlined, purpose-built runtime, Lightpanda empowers users to perform complex data extraction and agentic tasks with a much smaller compute footprint.
7. volcengine/OpenViking Not new today
OpenViking is an open-source context database designed specifically for AI Agents(such as openclaw). OpenViking unifies the management of context (memory, resources, and skills) that Agents need through a file system paradigm, enabling hierarchical context delivery and self-evolving.
First seen: March 15, 2026 | Streak: 3d
Analysis
OpenViking is an open-source context database specifically engineered to streamline the management of memory, skills, and resources for AI Agents. By moving away from traditional, fragmented vector storage, it adopts a hierarchical file system paradigm that allows developers to organize agent context into structured directories. Technically, the framework utilizes a tiered loading architecture (L0/L1/L2) to optimize token consumption and implements recursive retrieval and visual tracking, which enables developers to debug the agent's decision-making process with greater transparency.
Developers building complex, long-running AI Agents will significantly benefit from OpenViking, as it mitigates the common challenges of context loss and fragmented information retrieval. It is currently gaining traction because it transforms the chaotic nature of agent memory into a manageable, self-evolving system that mirrors local file organization. By offering a unified, observable, and cost-efficient way to handle large-scale context, the project empowers creators to build more intelligent and reliable agents that learn and improve over time.
8. shareAI-lab/learn-claude-code Not new today
Bash is all you need - A nano Claude Code–like agent, built from 0 to 1
First seen: March 09, 2026 | Streak: 2d
Analysis
The `learn-claude-code` repository provides a comprehensive educational framework for building an autonomous AI coding agent from the ground up, mirroring the functionality of Claude Code. Through 12 progressive sessions, it demonstrates how to build an agent loop that integrates tool usage, task planning, sub-agent delegation, and environment isolation. Technically, it relies on a minimalist "Agent Pattern" where an LLM orchestrates tool calls via a standard loop, layering on sophisticated mechanisms like background processing, task graphs, and persistent mailboxes without complicating the core execution logic.
This project is an ideal resource for software engineers and AI developers who want to demystify agentic workflows and move beyond simple chat-based interactions. It is currently trending because it offers a practical, "code-first" approach to understanding agent internals, providing mental models and clean reference implementations that are far more accessible than production-grade frameworks. By deconstructing complex concepts like worktree isolation and multi-agent coordination, it serves as a foundational guide for those looking to build their own custom CLI agents or embed agentic capabilities into their own software products.
9. p-e-w/heretic Not new today
Fully automatic censorship removal for language models
First seen: February 08, 2026 | Streak: 3d
Analysis
Heretic is an automated tool designed to remove "safety alignment" or censorship from transformer-based language models without the need for resource-intensive fine-tuning. It utilizes a sophisticated directional ablation technique—commonly referred to as "abliteration"—which is optimized via the Optuna framework to balance the suppression of model refusals with the preservation of overall intelligence. By co-minimizing KL divergence and refusal rates, the software autonomously identifies optimal parameters to modify model internals, making advanced intervention accessible to users without deep technical expertise.
This project primarily benefits researchers, AI hobbyists, and developers who seek to unlock the full potential of language models by bypassing rigid, often restrictive, built-in guardrails. It is gaining traction because it provides a user-friendly, high-performance alternative to manual, trial-and-error editing methods, consistently yielding models that retain logical integrity while responding to sensitive prompts. Additionally, Heretic offers unique research-oriented features, such as residual geometry visualization and analysis, making it a valuable instrument for those interested in the interpretability and inner workings of LLMs.
Agent harness built with LangChain and LangGraph. Equipped with a planning tool, a filesystem backend, and the ability to spawn subagents - well-equipped to handle complex agentic tasks.
First seen: March 17, 2026 | Streak: 1d
Analysis
Deep Agents is an opinionated, production-ready agent harness built on the LangChain and LangGraph ecosystems, designed to provide developers with a functional, ready-to-run autonomous agent out of the box. Technically, it functions as a compiled LangGraph application that integrates essential capabilities like task planning, filesystem interaction, sandboxed shell execution, and recursive sub-agent delegation. By handling complex context management—such as auto-summarization and file-based state tracking—the framework allows developers to focus on high-level customization rather than low-level wiring of prompts and tools.
This project is primarily intended for developers and AI engineers who need to deploy robust, agentic workflows rapidly without building infrastructure from scratch. It is gaining traction because it offers a "batteries-included" approach that remains fully extensible, provider-agnostic, and compatible with advanced LangGraph features like checkpointing and streaming. By providing a secure, MIT-licensed foundation that can be adapted for diverse tasks, Deep Agents serves as an ideal starting point for those looking to move beyond simple LLM wrappers toward building complex, multi-step autonomous systems.