Repository index
1016 AI, ML and Web3 projects · page 9 of 11
- SAASMAIL: EMAIL ON CLOUDFLARE WORKERS
Self-hosted email server for the SaaS teams, deployed on Cloudflare Workers. Supports WebMCP protocol enabling email integration directly into your apps. Written in TypeScript and uses the global Cloudflare network for instant delivery.
- PROFILER FOR AI AGENTS AND LLM
Graphsignal Profiler is a tool for debugging and monitoring inference. It helps find performance bottlenecks in deep neural networks and AI agents. Perfect for working with HuggingFace models.
- OMNICOREAGENT: HARNESS FOR PRODUCTION AI AGENTS
Open Python harness for production AI agents. Supports tools, MCP, memory, workspace, telemetry. Includes subagents, background tasks, and OmniServe APIs. 245 stars.
- CLONE VIRAL VIDEOS WITH AI
hypit lets you clone any viral video using AI‑agents. The system swaps the face, text, and B‑roll, and with one command outputs 100 variants ready for publishing.
- CREATE A GAME WITH ONE PROMPT
Noobi.ai — a local agent that turns a prompt into a vetted browser game via Codex App Server. Built with TypeScript + Electron, runs without a server.
- REFO_SKILL RESEARCH FOR AI AGENTS
refero_skill — a skill for AI agents based on a research-first approach. Provides access to 150K+ real app screens and flows via Refero MCP. Implemented in Python and supports integration with Claude, Codex and Cursor.
- HERMES ANDROID: CHAT WITH AI
Hermes Android is a Flutter/Dart client for chatting with Hermes Agent directly from your phone or tablet. Connect via local Wi‑Fi or a secure Tailscale tunnel and continue the chat with the AI assistant anywhere.
- AI GENERATES UI IN FIGMA
AI draws interfaces directly on the Figma canvas via JavaScript and returns existing designs as structured data. Supports Claude Code, Cursor, VS Code, and Windsurf without needing an API key.
- DEEPDIVE DEEP RESEARCH AGENT..
Deepdive — skill for Claude Code: 12‑stage pipeline with plan‑review, parallel search, claim verification with dissent protection and multilayer verification. Has 105 blocks, 29 channels, 460+ stat sources, 47 APIs and 1072 endpoints............
- USPEC: DESIGN SYSTEM DOCUMENTATION
uSpec generates design system documentation straight from your AI agent. The output can be exported to Figma or saved as a portable .md file. Integration works with Cursor, Claude Code, and Codex.
- PLATE AI TEMPLATE FOR EDITOR
Plate AI editor template on React 19, Next.js 16, Tailwind 4 and MCP. Includes TypeScript, ready integration of shadcn-ui and Radix UI for quick start. Allows customizing the toolbar and saving content to any storage.
- COMPLETE GUIDE CLAUDE COWORK FOR
The claude-cowork-guide repo offers full docs for Claude Cowork: 43 ready-made workflows and 70 effective prompts for knowledge workers and small businesses. All examples are Shell scripts that integrate with the Anthropic Claude API.
- POSTHORSE: CONTEXT WITHOUT SUMMARY
Posthorse provides native context windows without summarization for the fitchmultz/pi fork of the Pi coding agent enabling rollover tools, durable notes, and history recovery
- NAC: HARNESS FOR AGENTS AND TASKS
nac — open-source harness in Rust for executing long-running tasks of AI agents. It uses a central orchestrator, threads, and structured episodes to keep work aligned with the original intent.
- ONE CONTEXT ANY AGENTS
ACRYL creates a single persistent workspace and canonical context available to any coding agents. The project implements Agent Context Relay Yielding Lifecycles.
- HEADLESS ANDROID DEBUGGER FOR AI
Autonomous headless Android debugger for AI agents. Inspects runtime memory, sets breakpoints, and debugs live apps without UI. Written in Kotlin, 237 stars on GitHub.
- ALPHASUITE: OPEN PLATFORM FOR TRADING
AlphaSuite — an open‑source platform for quantitative analytics that lets you create, test and deploy professional‑grade trading strategies in Python. It supports backtesting, machine learning and LangChain integration.
- PENTEST HARNESS FOR HACKERS
Pentest Harness — a self‑hosted AI‑agent framework for authorized pentests, bug bounty, security labs, and CTFs. Connect your own model via API; all sessions run strictly locally.
- UNIVERSAL MCP SERVER FOR THE UNSPLASH
MCP server for searching Unsplash images via a simple HTTP API(REST). Supports filtering by image keywords, orientation, and result count, returning URLs and metadata in JSON format.
- TEXT CLASSIFICATION WITHOUT KEY
Zero-shot text classification over plain HTTP — no API key, no account. One Cloudflare Worker, CLI and MCP server work together. Try it now: https://classifier.dev
- PERSISTENT MEMORY FOR AI
Project persistent-ai-memory creates a local embedding store for AI assistants, LLMs, and GitHub Copilot inside VS Code. It retains context between sessions, enabling the model to recall past dialogues and code.
- QANAT ENGINE ALPHA DAG
Qanat — agent‑native engine for creating and backtesting alpha‑strategies as directed acyclic graphs. It uses DuckDB for fast data storage and MCP for integration with LLM agents. Allows describing research pipelines as code and launching them in a single click.
- PADSPANHA BLE ROOM TRACKING
padspanHA — the most comprehensive BLE room‑presence tracking system for Home Assistant. It provides room-level tracking, 3D floor plans, calibration, follow mode, and 24 view options.
- SICO — NEXT-GENERATION DIGITAL WORKER
Open Digital Worker platform for reliable task execution, continuous co-evolution of AI agents, and building Enterprise AI solutions.
- SLOTSTREAM LAUNCHES QWEN3.8 ON MAC
Run Qwen3.8‑Flash‑Next (125B MoE, 104 GB at 4‑bit) on Mac by streaming experts from SSD on demand. Implemented with MLX + Swift, it offers an Ollama‑compatible API.
- ONTOSPHERE – BROWSER RDF EDITOR
Ontology editor in the browser: load RDF from files, URL or SPARQL, draw nodes and edges, run OWL‑RL reasoning and apply various layout algorithms with clustering. Everything works fully client‑side, no backend required, and AI capabilities are available via the MCP interface.
- NAVIGATOR: CONTEXT FOR CLAUDE CODE
Navigator extends Claude Code sessions to 20+ exchanges, preventing crashes during long dialogues. It implements incremental context storage and a simple plug‑in integration, letting the AI retain all needed information.
- UNIVERSAL SKILL CATALOG FOR RESEARCH
Universal SKILL.md catalog for workflows: literature review, planning, project memory, writing, cross‑agent delegation for Claude Code, Codex, Gemini, Cursor, and other LLM clients. Supports Python and works with any LLM‑based agents.
- SWE-BENCH LIVE: EVALUATE LLMS IN CODING
SWE-bench Live — benchmark for evaluating AI agents on real GitHub issues. Automated code verification and comparison of any models.
- NAVIGATOR: CONTEXT FOR CLAUDE CODE
Navigator is a context engineering tool for Claude Code that keeps dialogue sessions alive for over 20 exchanges instead of crashing at the seventh. It preserves history and context, enabling the AI to finish started tasks without losing the thread.
- AGENT-AUDIT SECURITY SCANNER
Static security scanner for LLM agents. Detects prompt injection, audits MCP configs, and performs taint analysis. Includes 51 rules mapped to OWASP Agentic Top 10 (2026) and supports LangChain, CrewAI, AutoGen. Via CLI utility.
- PI‑FABRIC: AGENT FOR RASPBERRY PI
Pi‑Fabric — programmable tool and agent runtime for Raspberry Pi in TypeScript. Allows creating, orchestrating, and launching multi‑agent systems with MCP and coding‑agent support.
- NAMI: UNIFIED WORKSPACE FOR AI AGENTS
Nami unites top AI agents — Claude Code, Codex, Gemini and others — in a single workspace on macOS. You can run them side by side, switch between models and get combined results without losing context.
- OPENPOST: SOCIAL POSTS ON AUTO
OpenPost — open-source tool for creating, adapting, scheduling and tracking social media posts. Supports self‑hosted deployment and cross‑posting, written in TypeScript with Golang modules. Already 228 stars on GitHub.
- AUTOLITH — SELF‑MODIFYING AI AGENT
Autolith — self‑rewriting AI agent in Common Lisp, changing its code at runtime. It interacts with LLM via API, generates and improves programs, works as a CLI agent and extends via ordinary Lisp libraries.
- WIRESHARK‑MCP: AI PACKET ANALYSIS
The project lets you upload a .pcap file and ask questions in plain language. Answers are based on tshark data, providing accurate network traffic analysis. Integration with MCP turns the AI assistant into a full‑featured packet analyzer.
- AKSHARE-ONE-MCP: CHINA STOCK DATA
MCP server provides access to Chinese stock market data via akshare-one. Allows retrieving quotes, financial statements and macroeconomic indicators in real time. Compatible with all MCP clients.
- OLLITERT LLM SERVER ANDROID
OlliteRT turns your Android smartphone into a local inference server compatible with the OpenAI API. It works completely offline, ensuring privacy and open-source code in Kotlin.
- PR-COCKPIT: LOCAL AI REVIEWER FOR PR
Local AI reviewer powered by Claude and Codex. Reviews code in isolated worktrees, GitHub comments appear only after your approval.
- MOBILECODE: AI FOR MOBILE DEVELOPMENT
mobilecode is a fork of Opencode that builds and instantly previews iOS and Android projects. Supports building via Xcode and Android Studio, integrated with AI agents for code generation.
- AGENT LAUNCHER: ONE WINDOW FOR AIAGENTS
Agent Launcher unites Claude Code, Codex and other coding-agent CLI in a single cross-platform desktop app. Configure the agent once and run it without a terminal. Supports Windows, macOS and Linux.
- SPARK-X2.5: AGENTIC MODEL ON-DEVICE
Spark-X2.5 is an open model series that boosts agentic capabilities directly on the device. It supports code generation, task planning, and working with Hugging Face.
- OPEN WORKSPACE FOR AI AGENTS
Busabase — open-source database and workspace for AI agents. Stores structured data, knowledge, and skills, allows running apps. Supports human-in-the-loop, runs locally and self-hosted.
- KUBETERM: KUBERNETES AT HAND
Kubeterm provides a graphical interface for managing Kubernetes clusters. Works on desktop and mobile devices, simplifying monitoring and configuration directly in the browser.
- TRANSFORM WORK INTO KNOWLEDGE
Project project-cairn turns coding work into reusable knowledge for Claude Code and Codex. It extracts structure, generates prompts, and updates the knowledge base automatically.
- DEVCODEX UNIFIED AI CODING
DevCodex — runtime for intent‑driven AI coding: unified context, skills, approvals, validation and handoffs between six AI hosts. Written in JavaScript, works with Claude‑Code, Codex, Cursor.
- COOP: ISOLATED VM FOR CODE
coop creates an isolated environment based on Firecracker/KVM for safe execution of Claude Code and OpenAI Codex. The user gets a clean VM container where CLI commands can be run without affecting the host.
- CODFLOWPLATFORMFORALGERIA!!!
codflow — open-source e-commerce platform with cash‑on‑delivery for Algeria, built on TypeScript and ready for AI‑agent integration. Uses Astro and Cloudflare Workers for fast edge hosting. On the GitHub it already has 221 stars!!!
- AXOLOTL: AI-AGENT FOR QA
Axolotl analyzes code changes, generates test plans and runs real browser tests. It delivers evidence-backed merge verdicts.
- CREATE YOUR AI TWIN
agent-me lets you turn your knowledge, memories, and decisions into an open, inspectable AI‑agent twin. The project is written in Python and uses a knowledge graph for storing and retrieving information.
- FDEOPS: SKILLS FOR AI CODERS
A set of skills for AI agents: code search, dependency analysis, snippet generation, and refactoring. Compatible with Claude, Codex, and Cursor, deployed as Forward-developed Engineering.
- UNIVERSAL EPISODIC MEMORY
Engrim — a local project SQLite memory for AI agents. Supports working with Antigravity, Claude Code, Cursor, Windsurf and Codex without cloud attachment.
- SPEAKOFLOW: OFFLINE DICTATION WITH AI
Offline dictation tool for Windows, macOS and Linux — an alternative to Wispr Flow. It includes an AI assistant that, on request, reads the screen and answers questions. Everything works locally, with no data sent to the cloud.
- ZJT AI DRAMA GENERATOR
ZJT is an open-source platform on Python that uses AI to automatically create short dramas. From script and storyboard to video synthesis — the entire process is hands-free. It supports Docker and a multi-agent architecture.
- LOCAL MCP SERVER FOR COMFYUI
The project provides a local MCP server that lets AI agents directly control ComfyUI and launch its workflows. It supports Python, integrates easily with comfy‑cli and other tools.
- RUSSIAN TEXT FOR AI AGENTS
ru-text evaluates Russian text for AI agents on a 0–10 scale, cleaning neuroslop and applying typography to GOST. It checks style and uses 2000+ linguistic atoms.
- AWESOME REMOTE MCP SERVERS
This is a collection of ready‑made remote MCP servers that can be quickly deployed in any infrastructure. The project simplifies MCP integration with LLM agents by providing ready‑to‑use configurations and startup scripts.
- AGENT METHODOLOGY FOR SOLO
Project bigpowers turns years of engineering experience into ready agent skills for solo developers. Automates routine via Bmad, Claude Code and Cursor Rules. 216 stars on GitHub.
- IMAGE GENERATION IN DEEPSEEK HARNESS
dsh-image-gen — a plugin for DeepSeek Harness that generates images from text prompts directly in the chat. Written in TypeScript, it supports style and parameter configuration via simple commands and integrates with Gemini and other LLMs.
- CLAUDISH: TRANSLATE ENGLISH AND CLAUDISH
Claudish - a small Python library that translates between English and Claudish using tiny ProgramAsWeights functions. With 215 stars on GitHub it demonstrates a simple NLP translator for experiments with languages and LLMs.
- EVOQUANT: AI AGENT FOR QUANTITATIVE RESEARCH
EvoQuant — an AI agent that performs the quantitative research loop: from hypothesis generation to strategy backtesting. It operates autonomously in Python, using a multi‑agent architecture to hunt for alpha.
- BANKMCP: READ YOUR BANK!
BankMCP™ — a self‑hosted read‑only MCP server that connects to bank accounts via Enable Banking (Open Banking). It lets AI models Claude and Ollama securely query balances and transactions, so operating as the standard MCP .
- EVALFORGE — LLM EVALUATION IN CI
EvalForge is a tool for LLM evaluation in CI/CD. Provides evaluator-neutral evidence, baseline regression gates, and JSON, JUnit, SARIF reports. Integrates with GitHub Actions.
- AGENTMAIL: EMAIL, SMS, CALLS FOR AI
Agenticmail provides an API for sending and receiving real email messages, SMS, and outbound voice calls from AI agents. It integrates with any LLM frameworks and supports programmable control over communication channels.
- MCP SERVER FOR FREECAD AI
Project provides MCP server and bridge for FreeCAD, allowing to connect AI agents and control geometry via natural language. Extend functionality via Python scripts, and everything works as an addition to Workbench.
- PYMEMORYEDITOR: WORKING WITH PROCESS MEMORY
PyMemoryEditor — a pure‑Python library for reading, modifying and searching in the memory of any running process. Just a few lines of code let you work with memory on Windows, Linux and macOS.
- VOICE AGENT FOR AI AGENTS
tapq is a voice agent that lets you communicate with Claude Code, Codex and other AI agents via microphone. You can respond to their prompts, give instructions and ask what they did, or just nod.
- PERSISTENT MEMORY FOR CLAUDE CODE
claude-db adds persistent memory to Claude Code, allowing dialogs to be stored in your own database. Supports any SQL or NoSQL storage via a simple TypeScript SDK. Integration takes just a few lines of code, with no external services.
- LEMMALOG: DATALOG FOR AGENT MEMORY
Lemmalog — a Datalog engine for LLM agent memory. It supports stratified rules, provenance‑tracked facts, incremental derivation, and an MCP server that lets you use the engine as a shared brain.
- CONTROL TIA PORTAL WITH ONE MCP
TIA_Portal_Openness_MCP controls Siemens TIA Portal V20/V21 (STEP 7, WinCC Unified) from any MCP client. It lets you create projects, generate PLC and HMI, compile and diagnose — without Openness code thanks to the pre‑built runtime.
- CLAIETOP: HTOP FOR CLAUDE SESSIONS
Monitor cost, cache, and Claude Code models in real time. Receive smart alerts for budget overruns and efficiency drops.
- MANAGE MULTIPLE PROJECTS
Grasp combines multiple projects and parallel AI‑agents into a single visual panel. You see dependencies, manage tasks, and speed up releases without switching context.
- SESSION-KNOWLEDGE LOCAL KNOWLEDGE BASE
The project stores Claude Code session history in a local SQLite database and provides 13 MCP tools for search. Indexing runs on pure Python standard library, and no data leaves your machine.
- AI CONTROLS ALTIUM VIA MCP
Open-source MCP server gives AI direct access to live Altium Designer, KiCad, or EasyEDA Pro sessions. 300+ tools for schematics, PCB, libraries, auto-placement, and panelization. Apache-2.0.
- NO_HUMAN: FULL AUTONOMOUS AI AGENT
Turns tickets into ready pull‑requests, covering the full cycle: analysis, code generation, testing, and review. Fully open‑source, runs locally on Python, with no cloud dependencies. Stars 212.
- PHI — UNIVERSAL AI CODER
PHI is a universal AI coding agent. Supports infinite LLM providers, sub-agents, and hashline edits. Permission gate controls file and command access.
- TOKENIZER BPE FOR OPENAI
Tokenizer — BPE tokenizer for TypeScript and .NET for OpenAI models. Trains vocab, encodes and decodes tokens, compatible with tiktoken. Easily integrates into AI products.
- CATALOG OF QA SKILLS FOR AI
QA Skills is a curated catalog of testing skills for AI coders like Claude Code, Cursor, and Copilot. Each skill is described as a ready‑to‑use module that agents can plug in to perform specific UI, API, and e2e.
- REAL USE CASES HERMES AGENT
awesome-hermes-usecases — curated collection of real-world use cases for Hermes Agent, the self‑improving AI agent from Nous Research. Each example is backed by primary sources and demonstrates practical automation scenarios.
- VICOA: AGENT IDE FOR CODE
Vicoa — agent IDE that lets you control a team of coding agents from any device: desktop PC, smartphone, or VPS. The project is open source, self-hostable, and written in Python.
- GUIDE TO AI ARCHITECTURE
awesome-ai-architect — curated guide list blending AI engineering with solution architecture. Speeds junior‑to‑senior AI‑architect growth via clear roadmaps, templates, interview prep. 200+ stars validate its utility.
- OPENAI TTS FOR HOME ASSISTANT
The openai_tts component for Home Assistant uses OpenAI Speech API (or any compatible endpoint) and outputs high‑quality speech. There is an optional chime before/after speech and audio normalization for consistent volume.
- PRYSAI PLAYBOOK: COMPLETE LLM GUIDE
LLM playbook on practice: 8 locales, unified core, Codex track and adapters for ChatGPT, Claude Code, Gemini, DeepSeek, Grok.
- CHIASMUS — MCP SERVER FOR FORMAL VERIFICATION
Chiasmus — an MCP server that gives language models access to formal verification tools. It converts queries into formal specifications and generates proofs using Coq, Isabelle, Lean via MCP & supports counterexample detections.
- FREE TPU FOR QWEN3.8-27B
The kaggle-tpu-lab project runs Qwen3.8-27B (bf16) on a free Kaggle TPU, providing an OpenAI‑compatible API. It supports a 262k token context and ~130 tokens/s speed, works with Claude Code, Codex, Opencode and Pi.
- ALETHEORE AUDIT REPOSITORY CLI
Aletheore — CLI scanner that deterministically checks the repository, starts an MCP server, shows a live dashboard, and via GitHub Action posts PR diffs.
- FIRSTCODER: LOCAL AI AGENT FOR CODE
FirstCoder — open-source local coding agent in Python. Useful enough to run, small enough to read end to end, and structured enough to learn from. A practical reference when building your own agent.
- REVIVA: LOCAL AI WORKSPACE
Reviva — local AI‑workspace: ask questions, take notes, review material and create content from your files. There is a Wiki‑KB with RAG, AI agents, customizable skills and content creation tools.
- CARWATCH AI AGENT CAR
CarWatch turns Raspberry Pi 5, a dashcam, and local AI into a chat agent for your car. Talk to the vehicle, get data, and control functions in real time. This is a CodeWatch analogue for the garage.
- OPENSOURCE MEMORY FOR AI AGENTS
ownmem — open-source, Git memory for AI coding agents. Provides local recall for Claude Code, Codex, Cursor, Gemini CLI and compatible tools. Indexes the repository and stores embeddings for fast search.
- MCP SERVER FOR 1C:ENTERPRISE
MCP server for 1C:Enterprise — AI assistant sees the configuration and generates accurate BSL code. One binary, zero dependencies, 11 ready tools.
- LOOMFLOW: LIGHTWEIGHT AI WORKFLOW BUILDER
LoomFlow is a lightweight AI-native workflow builder. Describe your idea in natural language, get a visual scheme and immediately publish it as an API. Install with one click via Docker.
- LOOMIC AI CANVAS FOR CREATIVITY
Loomic - open-source AI agent for infinite canvas. Generates images and video via chat, replacing Lovart, CapCut Video Studio and Canva AI. Supports TypeScript and extends via plugins.
- SLOPTRIM: AI-SLUDGE DETECTOR
Sloptrim evaluates agent’s text files and returns a 0‑100 score for AI‑generation probability. Works offline using pure Python — no network, no dependencies, no models. Fits Claude Code and any agents for checking AI‑slop.
- CELERITY-PANEL SELF-HOSTED PROXY
Web panel for self-hosting Hysteria 2 and Xray VLESS with cascade topology (Forward/Reverse Chain). Provides SSH setup, server groups, load balancing, ACL filter, API, webhooks, S3 backup, MCP and subscriptions for Clash/Sing‑box/Hiddify.
- FREEBUCKS PROXY: CODING GATEWAY
Multi-account OpenAI‑compatible gateway for coding models. Supports pooled keys, session lifecycle management, TLS stealth, and an embedded admin dashboard. Works without ads or CLI, accessible only via the /v1/chat/completions endpoint.
- ZOTSEEK AI SEARCH IN ZOTERO
ZotSeek adds semantic search to your Zotero library using vector embeddings. The built-in MCP server lets you connect Claude Code and Codex as AI agents for meaning‑based queries. Everything runs completely locally, with no data sent over the network.
- GEO‑SLEUTH DETERMINE PHOTO LOCATION
The agent locates a photo’s shooting spot using OpenStreetMap geometry, elevation profile, satellite imagery and Street View. It works with Claude Code, Codex, Cursor, Gemini CLI, OpenCode and GitHub Copilot, showing each step.
- MULTI-AGENT ALTERNATIVE TO CLAUDE TAG
AgentConnect — open-source multi-agent platform that lets you connect any AI‑agent to your workflows. They learn on the fly, working alongside your team, and support the Agent‑Client‑Protocol and Agent Sandbox.
- BUILD YOUR AI INTRANET
Autonomous grid unites your computers into a single infrastructure for running AI models: training, fine‑tuning and inference. Supports distributed computing, fine‑tuning via GRPO and operation in a home lab.