Packages

MCP server providing complete macOS desktop automation capabilities including mouse control, keyboard input, screenshot capture, and UI element interaction via Accessibility APIs.

Managed cloud vector database service built on the open-source Chroma engine, offering a REST API and Python/Node SDKs for storing and querying embeddings in agent RAG pipelines.

Extends Django to handle WebSockets, long-polling, background tasks, and other async protocols alongside standard HTTP. Django Channels replaces Django's WSGI server with ASGI (via Daphne or Uvicorn), adds channel layers (Redis-backed pub/sub for cross-process communication), and provides WebSocket Consumers (async or sync class-based handlers). Key concepts: Consumer (handles WebSocket connection lifecycle), Channel Layer (Redis pub/sub for broadcasting to groups), Channel Groups (broadcast to all connections in a group). Used for real-time agent status updates, live agent chat interfaces, streaming agent responses to browser, and collaborative agent workspaces.

FastAPI-inspired async Python framework for building message broker microservices. Single unified API for Kafka, RabbitMQ, NATS, and Redis Pub/Sub. Automatic schema generation, built-in testing with TestClient, and type-safe message handling with Pydantic. Dramatically reduces boilerplate for event-driven Python services compared to raw aiokafka or pika.

Frankfurter is a free, open-source, no-API-key-required currency exchange rate API that serves European Central Bank reference rates for 32 currencies with HTTPS and transparent ECB sourcing.

Google Sheets API v4 — read, write, format, and manage Google Sheets spreadsheets programmatically, enabling agents to use Sheets as a lightweight database or reporting surface.

Helicone LLM observability proxy — a drop-in proxy that intercepts LLM API calls (OpenAI, Anthropic, etc.) to log requests/responses, track costs, enable caching, rate limiting, and provide analytics dashboards with minimal code changes.

Email finder and verification API that discovers professional email addresses by domain or name, verifies deliverability, checks for disposable or catch-all configurations, and returns confidence scores. Used for B2B sales prospecting and outreach automation.

Provides a command-line suite and library API for creating, editing, converting, and compositing raster and vector images across 200+ formats including JPEG, PNG, TIFF, WebP, HEIC, PDF, and SVG.

MCP server for Korea's G2B (Government to Business) procurement system — the Nara Market platform for Korean government contracts and procurement. Enables AI agents to search and retrieve Korean government procurement opportunities, contract data, and bidding information.

Original JavaScript monorepo management tool, now maintained by Nx and integrated with Nx task scheduling. Manages multiple npm packages in a single repository — versioning, changelogs, and publishing to npm. Lerna v6+ integrates with Nx for task running and caching. Known for its two versioning modes: fixed (all packages same version) and independent (packages version independently). Still widely used but many teams have migrated to Turborepo or Nx.

LLM-powered PDF and document parsing service from LlamaIndex. LlamaParse converts complex PDFs (multi-column, tables, charts, images) into clean Markdown or structured text optimized for LLM ingestion and RAG. Uses LLMs to understand document structure rather than pure text extraction — producing better table formatting, section hierarchy, and figure descriptions. Designed as the ingestion layer for LlamaIndex RAG pipelines but usable independently.

Provides 132 domain-specific skills (structured documentation with acceptance criteria) for AI coding agents working with Azure services, covering Python, .NET, TypeScript, Java, and Rust SDKs.

Async Python MongoDB driver — the official asyncio-native MongoDB driver for Python, wrapping PyMongo for async/await usage. Motor features: AsyncIOMotorClient, AsyncIOMotorDatabase, AsyncIOMotorCollection, async CRUD (await collection.insert_one(), await collection.find_one()), async aggregation pipeline, change streams with async_for, GridFS async, connection pooling, Tornado support, and full PyMongo feature parity via async interface. Required for FastAPI, aiohttp, and asyncio-based agent backends that use MongoDB.

Production ML inference server from NVIDIA optimized for GPU workloads. Supports TensorFlow, PyTorch, ONNX Runtime, TensorRT, and Python backends. Provides HTTP and gRPC APIs for inference, model management, and health checking. Features dynamic batching, model ensembles, concurrent model execution, and performance analysis. Industry standard for high-throughput GPU inference in production.

Declarative configuration management tool where agents pull compiled catalogs from a Puppet Server, applying manifests that describe desired system state using Facter facts and Hiera data hierarchies.

Redis Cloud is managed Redis-as-a-service by Redis, Inc., providing the blazing-fast in-memory data store for caching, session management, pub/sub messaging, rate limiting, distributed locks, and queue patterns. The Redis Cloud REST Management API (redislabs.com) handles the control plane: provisioning databases, configuring eviction policies, setting memory limits, and managing subscriptions. This is completely separate from the Redis data protocol (RESP) used for actual data operations. Agents interact with the management plane via REST and the data plane via standard Redis clients.

A production-grade multi-agent orchestration platform built on Rust/WebAssembly that deploys coordinated swarms of 60+ specialized AI agents within Claude Code, with persistent memory, cost optimization, and Byzantine fault-tolerant consensus.

SendGrid Marketing Campaigns API — contact management, list segmentation, and email campaign sending for marketing use cases, distinct from SendGrid's transactional Mail Send API.

REST API that returns structured JSON from Google, Bing, YouTube, Amazon, and other search engines without managing headless browsers.