Packages

An open-source toolkit that connects AI assistants to civic/public data sources via the Model Context Protocol (MCP). It includes (or wires up) MCP servers for Socrata-based NYC Open Data (e.g., 311, inspections, housing) and Google Data Commons, enabling natural-language querying and generation of analyses/visualizations against live civic datasets.

Graph-based RAG system that parses multi-language codebases using Tree-sitter, builds knowledge graphs in Memgraph, and enables natural language queries about code structure and relationships. Supports parsing 11+ languages, AI-powered Cypher query generation, surgical code editing with AST targeting, shell command execution, and real-time graph updates via file watching. Includes an MCP server for Claude Code integration.

MCP server that enables AI assistants to interact with multiple databases simultaneously through a unified interface. Dynamically generates per-connection tools for querying, executing statements, managing transactions, exploring schemas, and analyzing performance across MySQL, PostgreSQL, SQLite, Oracle, and TimescaleDB.

Provides an MCP server for Industrial IoT/SCADA/PLC that unifies MQTT sensors/actuators and Modbus TCP/RTU device operations, with optional time-series storage (InfluxDB) and caching (Redis). Exposes MCP tools via HTTP endpoints using FastAPI (with /docs, /health, /metrics) for agent orchestration (e.g., PolyMCP). Includes a simulation mode for running without external hardware.

Ludus FastMCP is a Python-based Model Context Protocol (MCP) server that automates Ludus cyber range (v1 and v2) management via a large set of MCP tools (190+). It supports range lifecycle tasks (create/configure/deploy/manage), templates and YAML generation, scenario deployment, v2 blueprints, v2 groups, diagnostics/migration (v2), and SIEM/security monitoring integrations.

An MCP (Model Context Protocol) server that provides a single tool to look up authoritative Svelte 5/SvelteKit definitions extracted from TypeScript declarations, with caching and fuzzy matching for identifier queries.

Subcog is a persistent memory system for AI coding assistants. It captures decisions/learned context during coding sessions, stores them in SQLite (default) plus indexing (FTS5) and vector search (usearch HNSW), supports hybrid retrieval (BM25 + vector with RRF fusion), provides a knowledge-graph layer, and exposes an MCP server for agent interoperability. It also offers optional HTTP serving with JWT auth and Claude Code hook integration.

XHS-Downloader extracts and downloads images, videos, and metadata from Xiaohongshu (RedNote/小红书) posts, profiles, and search results, with support for TUI, REST API, MCP server, and browser UserScript interfaces.

An open-source workflow automation platform (Zapier alternative) with 280+ integrations, a visual no-code builder, AI agent capabilities, and MCP server support that exposes all pieces as MCP-compatible tools.

Async video messaging platform API for managing, sharing, and embedding screen recordings and video messages within team workflows.

An open-source platform for building and deploying fully-managed AI agents and long-running workflows in TypeScript, with durable task execution, automatic retries, queues, human-in-the-loop pauses, and full observability.

Aguara is a local security scanner (CLI + library + Docker) that statically analyzes AI agent skills and MCP server/tool configurations to detect issues such as prompt injection, data exfiltration, credential leaks, and supply-chain/persistence/rug-pull patterns. It supports decoded/obfuscated content scanning, cross-file taint/risk flow analysis within an MCP server directory, and produces terminal/JSON/SARIF/Markdown outputs suitable for CI.

Allyson is an MCP (Model Context Protocol) server that provides AI-powered SVG animation generation. It transforms static image files (PNG, JPG, SVG) into animated SVG components by accepting natural language animation prompts and outputting React-compatible TSX files with Framer Motion animations.

Provides an MCP stdio server that automates AutoCAD LT (Windows) via AutoLISP dispatch using a File IPC backend, and alternatively generates/render DXF headlessly via an ezdxf backend. Exposes consolidated MCP tools for drawing/entity/layer/block/annotation/P&ID/view/system operations, plus an execute_lisp escape hatch (File IPC only).

Starter template for building an MCP Apps (Model Context Protocol) application with a Node.js/Express MCP server and React-based widget resources. Includes an example “echo” tool with Zod input validation, resource/widget registration, UI capability negotiation, inline/PIP/fullscreen display mode support, and local dev/testing tooling (Storybook, Vitest) plus Docker support. Also documents how to expose the MCP endpoint publicly for hosts such as ChatGPT via Pomerium SSH tunnel and how to connect via the host’s connector UI.

An MCP server that provides versioned, hot-reloadable Claude prompt templates (YAML), validation rules ("gates"), reasoning frameworks ("methodologies"), and response styles, plus a workflow/operator syntax to chain steps, run verification loops, and (optionally) hand off mid-chain steps to subagents with shared context.

ClawMem is a local-first context engine/memory for AI coding agents. It indexes markdown/project documents into a SQLite-based “vault”, performs hybrid retrieval (BM25 + vector + reranking + graph/intent-based traversal), and integrates with Claude Code via hooks + an MCP stdio server and with OpenClaw via a ContextEngine plugin or MCP tools. It also captures session decisions/handoffs and writes them back to the same local vault; it includes measures against prompt injection in surfaced content and supports deduplication, lifecycle (pin/snooze/archive), and crash-safe transaction-safe indexing.

Provides a local Model Context Protocol (MCP) server that exposes filesystem operations (list/read/write-like operations, mkdir/delete/copy/move, info, search, edit, and list_allowed_dirs). All operations are intended to be constrained to configured allowed absolute directories.

Provides an MCP (Model Context Protocol) server for Membrane that exposes Membrane integration actions as MCP tools. Supports connecting over HTTP streamable transport and (deprecated) SSE, with token-based authentication and optional static vs dynamic tool exposure.

Provides a Python CLI (`kd`) and an MCP server to convert video URLs (YouTube/Bilibili/Facebook) into structured knowledge outputs by extracting subtitles when available, otherwise performing local ASR (Qwen3-ASR on Apple Silicon), and optionally generating summaries via external LLM providers (Google/OpenAI/Anthropic).