Packages

Elasticsearch REST API — distributed search and analytics engine with full-text search, vector search (kNN), aggregations, and real-time indexing, enabling agents to build powerful search experiences, log analytics, and hybrid semantic search pipelines.

Comprehensive REST and GraphQL API for GitLab's integrated DevOps platform, enabling programmatic control over CI/CD pipelines, jobs, repositories, merge requests, and the full software development lifecycle.

Battle-hardened open-source load balancer and proxy for TCP and HTTP traffic. HAProxy is renowned for its reliability, performance, and rich health-checking capabilities. Used in production by GitHub, Stack Overflow, and many high-scale deployments for agent backend load balancing, health check routing, and TCP proxy for non-HTTP agent protocols.

Persistent immutable data structures for JavaScript — List, Map, OrderedMap, Set, Stack, Record. Immutable.js uses structural sharing so operations that 'modify' data return new versions that share unchanged subtrees, making immutability memory-efficient. Common in Redux state management and React applications where reference equality enables fast change detection.

Full-text search abstraction for Laravel Eloquent models — adds search driver via Searchable trait, indexing, and search() method with driver-agnostic API. Scout features: use Searchable in Eloquent model enables automatic syncing (queue-based), Agent::search('query')->get() for full-text search, toSearchableArray() customizes indexed data, searchable()/unsearchable() for manual index management, soft delete support, custom search engines (Algolia, Meilisearch, Typesense, Elasticsearch via community drivers, database engine for development). Automatic reindexing on model events (save/delete). Driver-agnostic search with pluggable backends — switch from Algolia to Meilisearch without changing search code.

Asynchronous coordination layer for AI coding agents providing persistent identities, inboxes/outboxes, searchable message threads, and advisory file reservations (leases) to prevent conflicts. Git-backed for auditability, SQLite-indexed for queries.

MCP server providing Open Source Intelligence (OSINT) capabilities to AI agents — enabling agents to gather publicly available information about individuals, organizations, domains, IP addresses, and digital assets. Integrates OSINT tools and techniques into agent-driven threat intelligence and security research workflows.

Turns Python functions into LLM calls via a @prompt decorator, using type annotations to automatically parse and validate structured outputs with minimal boilerplate.

Mambu Cloud Banking REST API for composable banking-as-a-service platform. Enables AI agents to manage loan origination and servicing lifecycle automation, handle deposit account creation and transaction management, access credit product configuration and eligibility management, retrieve loan repayment schedule and amortization data, manage customer and account portfolio management, handle interest and fee calculation and accrual management, access disbursement and repayment transaction processing, retrieve portfolio analytics and loan performance data, manage accounting and GL transaction generation, and integrate Mambu core banking with fintech apps, payment processors, and compliance systems.

Drop-in React performance optimizer. Million.js replaces React's virtual DOM diffing algorithm with a hyper-optimized 'block virtual DOM' that's up to 70% faster for rendering. Works as a Babel/SWC compiler plugin — automatically wraps eligible React components in Million's faster renderer. No code changes required; install the package, add the plugin, and eligible components render faster. Used to speed up existing React applications without rewriting.

Embeddable persistent document store for Node.js and browser with a MongoDB-compatible API. Stores data in a file-based append-only format (or in-memory for tests). No external server required — the DB runs in-process. @seald-io/nedb is the actively maintained fork of the original archived nedb project.

Python library for structured text generation using LLMs with constrained decoding. Outlines guarantees LLM output matches a JSON schema, regex pattern, Pydantic model, or grammar by manipulating the token generation process — not just post-processing. Works with local models (transformers, llama.cpp) and APIs. Eliminates LLM output parsing failures by making invalid outputs structurally impossible.

Pure JavaScript PDF document generation library for Node.js and browser. PDFKit creates PDF documents programmatically — text, images, shapes, vector graphics, tables (via manual positioning), fonts, and hyperlinks. Outputs as a readable stream. Used for invoice generation, report PDFs, and any agent workflow that needs to produce PDF documents without conversion from HTML.

TypeScript-first GraphQL schema builder with a plugin system. Define GraphQL schemas using TypeScript code — no SDL files. Provides excellent type inference where resolver args and return types are inferred from schema definition. Plugin ecosystem includes: Relay, Prisma, Drizzle, Auth, Complexity, DataLoader, Errors, and more. Used as the code-first alternative to SDL-first tools like nexus or type-graphql.

Low-level rich text editor toolkit for the web. Provides a robust document model, schema definition, editing operations, and an extensible plugin system. Used as the foundation for Tiptap, Remirror, Atlassian's editor, and The New York Times's editor. Not a ready-to-use editor — it's building blocks for creating custom editors. Enables collaborative editing (via Y.js or similar), custom schemas, and highly specialized document editors.

Official Python SDK for Sentry — the error tracking and application monitoring platform. One sentry_sdk.init() call adds automatic exception capturing, performance tracing, distributed trace propagation, and integrations for Django, Flask, FastAPI, SQLAlchemy, Redis, Celery, and more. Captures agent errors and performance data and sends them to Sentry Cloud or self-hosted Sentry.

The most widely-used JavaScript library for reading and writing Excel files (XLSX, XLS, ODS, CSV, and 30+ spreadsheet formats). Works in both Node.js and browsers without native dependencies. Parses spreadsheet data to JavaScript arrays/objects and serializes JS data to spreadsheet files. Used for Excel report generation, data import from Excel uploads, and spreadsheet format conversion. The xlsx npm package is the Community Edition; SheetJS Pro adds advanced features.

Universal UI library and design system for React and React Native. Write components once and render on iOS, Android, and web with native performance. Includes a compiler that evaluates styles at build time (converting dynamic styles to static CSS variables). Ships with 100+ components, theming system, and animations. Positioned as the solution to 'style once, run everywhere' — replacing the need for separate web and native component libraries.

Proxy-based state management library for React by Poimandres (pmndrs — also behind Zustand, Jotai, React-Spring). Valtio makes JavaScript objects reactive using Proxy — mutate state like plain JavaScript objects, and React components auto-subscribe to changes. Very little boilerplate — create a store with proxy(), mutate it directly, and useSnapshot() in components for reactive access. Mental model closest to MobX but lighter and proxy-based.

Fast MySQL client for Node.js with async/await support, prepared statements, and connection pooling. The standard MySQL driver for Node.js — replaces the older mysql package with better performance, native Promise support, and TypeScript types. Supports MySQL 5.7+, MySQL 8.0+, MariaDB, PlanetScale, and other MySQL-compatible databases. Used extensively with ORM frameworks like Prisma, Drizzle, Sequelize, and TypeORM as the MySQL transport layer.