Packages

An MCP (Model Context Protocol) server that provides Kubernetes diagnostic tooling for pods and clusters, including pod health analysis, cluster health overviews, pod log analysis for error patterns, and workload recommendations, backed by Kubernetes API access via in-cluster service account or kubeconfig.

Provides a Go-based Model Context Protocol (MCP) server that exposes Kubernetes resources as MCP tools (e.g., Pod/Deployment/Service/etc.) and also offers a generic custom-resource tool for CRDs. Operations are executed against a Kubernetes cluster using a provided kubeconfig.

LitServe is a lightweight Python framework (FastAPI-based) for building and serving custom AI inference servers. Users implement a LitAPI with `setup()` and `predict()` (and potentially more advanced logic like batching/streaming/routing), then run it via `LitServer` to expose an HTTP API for inference pipelines, including agents, RAG, and multi-model workflows. It supports self-hosting and deployment via Lightning’s cloud offering.

MCP-Bench is an evaluation framework that benchmarks tool-using LLM agents on complex tasks using the Model Context Protocol (MCP). It orchestrates discovery/connection to multiple MCP servers, runs benchmark tasks (single- and multi-server settings), and evaluates outputs (including LLM-as-judge).

Provides a deployable MCP Gateway (reverse-proxy via Nginx) and a MCP Registry/UI to register and govern MCP servers, discover tools (including schemas via ListTools), and route MCP client traffic to enabled backend MCP servers using per-service URL paths. Includes health checks and real-time UI updates via WebSockets.

mcphub is a Python package/CLI that helps developers configure, install/run, and connect MCP (Model Context Protocol) servers into AI applications. It supports stdio-based MCP connections and optionally an SSE-based mode (via a `mcphub run ... --sse` supergateway-style setup), provides a JSON `.mcphub.json` configuration format, and includes adapters to integrate MCP tools with frameworks such as OpenAI Agents, LangChain, and Autogen.

An MCP (Model Context Protocol) server that lets an AI assistant trigger cross-platform desktop notifications on behalf of the user. It exposes tools such as basic/urgent notifications, notifications with icons and timeouts, and a richer fully customizable notification.

Provides an MCP server that fronts a Prometheus instance so an MCP-capable client (e.g., Cursor) can query Prometheus for metrics, labels, targets, rules, and alert information using an API key for access control.

PencilPlaybook is a Claude Code “skill” intended to guide Claude in producing UI/design outputs inside Pencil.dev by embedding perceptual design defaults (token lookup tables), structured workflows for working with .pen canvases, and presets/scaffolds for common screen archetypes. It also references a setup wizard that configures design-system presets and claims to provide a complete parameter reference for 12 Pencil MCP tools.

Qurio is a self-hosted ingestion and retrieval (RAG) engine for AI coding assistants. It crawls or ingests local documents (e.g., web pages, PDFs, Markdown), chunks content structurally, embeds it (Gemini for embeddings), stores vectors/metadata in Weaviate/PostgreSQL, and exposes retrieval to agents via an MCP server over a JSON-RPC 2.0 endpoint.

Sample Python project showing how to build an agentic web automation assistant using AWS Bedrock (Claude and Amazon Nova) with tool use, human-in-the-loop interruptions, and vision (screenshots). Includes progressive steps culminating in an MCP (Model Context Protocol) client/server refactor and conversation history/token-management enhancements.

Solana MCP server that exposes a set of Solana JSON-RPC-like capabilities (accounts, blocks, transactions, tokens, cluster/system info, staking/governance) to an MCP-capable client (described as working through Cline) using Rust.

Provides a local, read-only MCP server that exposes Square data as queryable tables/columns via a wrapped CData JDBC Driver. It runs as an MCP stdio server intended for use with local MCP clients (e.g., Claude Desktop) to retrieve live Square data without requiring users to write SQL.

zin-mcp-client is a local CLI/Web UI MCP client that launches and proxies STDIO-based MCP servers (via a local MCP config), and connects them to local LLMs running on Ollama to perform tool-calling through the MCP servers. It also provides an MCP proxy for integration with Open WebUI.

Arduino MCP server (Go) that exposes MCP tools to list connected Arduino boards and compile/upload Arduino sketches, intended to be run locally via stdio (e.g., from VS Code MCP settings).

Provides a headless MCP server for Binary Ninja, exposing a large tool surface (analysis, disassembly/IL, patching, types/metadata, undo/redo, memory operations, scripting, and more) over stdio and TCP transports, including a fake backend mode for CI/dev without a Binary Ninja license.

A Python utility for converting Compiled HTML Help (CHM) files to Markdown format, specifically optimized for Revit API documentation. This tool extracts HTML files from CHM documents and converts them to well-formatted Markdown files, making technical documentation more accessible, version control friendly, and AI-readable.

A local, read-only MCP server that exposes DocuSign data via CData JDBC Driver (SQL-modeled tables/columns) to LLM clients (e.g., Claude Desktop) over MCP stdio. It provides MCP tools to list tables/columns and run SQL SELECT queries through the JDBC driver.

Provides an MCP (Model Context Protocol) server (implemented with Spring AI MCP) that exposes Elasticsearch operations as MCP tools. It supports cluster health/statistics, CAT endpoints for mappings/indices/aliases, shard allocation queries, and document search against Elasticsearch indexes (including an AI-generated queryDSL workflow).

A production-oriented Next.js/TypeScript template for building LangGraph.js-based AI agents with dynamic tool loading via Model Context Protocol (MCP), optional human-in-the-loop tool approval, persistent thread-based conversation memory backed by PostgreSQL (LangGraph checkpointer), and real-time streaming responses via SSE. It also includes multimodal file upload/storage using S3-compatible backends (e.g., MinIO for dev).