Skyvern automates browser-based workflows using LLMs and computer vision, providing a Playwright-compatible SDK and no-code workflow builder that can interact with any website without brittle XPath-based scripts.
Converts documentation sites, GitHub repos, PDFs, videos, and 10+ other source types into structured knowledge assets that can be exported to Claude skills, LangChain, LlamaIndex, Cursor, and other AI systems with a single command.
An MCP server that reduces context window consumption by sandboxing raw tool output (98% reduction) and maintains session continuity across conversation compactions using SQLite-backed event tracking with FTS5/BM25 search.
Context Mode is an MCP server that intercepts large tool outputs (test results, logs, docs, web snapshots) and returns only concise summaries to Claude, reducing context consumption by up to 98% and extending session longevity.
Dynamically manages and transforms context metadata across distributed systems, enabling seamless information exchange and semantic interoperability.
Semiotic is a React data-visualization library designed for AI-assisted development — 30+ chart types (XY, categorical, network, geo, realtime, coordinated views) with a `semiotic-mcp` MCP server exposing `renderChart`, `getSchema`, `suggestChart`, `diagnoseConfig`, and `reportIssue` tools so coding assistants generate correct chart code on the first try.
MCP server that lets multiple Claude Code instances find each other and exchange messages instantly — when you're running several sessions across different projects, any Claude can discover the others and send messages that arrive as `<channel>` events in real time.
High-performance data-processing engine built on a multithreaded C++ core with 50+ Python-extensible pipeline nodes (13 LLM providers, 8 vector databases, OCR, NER, PII anonymization) — exposed to AI assistants through TypeScript, Python, and MCP SDKs so agents can build, run, and deploy visual `.pipe` workflows on Docker, on-prem, or RocketRide Cloud.
Indexes codebases using tree-sitter AST parsing and lets AI agents retrieve specific symbols with byte-level precision, reducing code-reading token costs by up to 99%.
AI-native SwiftUI component library with an MCP recipe server — production-ready animations (Shimmer, Typewriter, OrbitingLogos), charts (Line/Bar/Donut/Radar/Heatmap), UI components (Alert, Onboarding, Stepper), and multi-file modules (SWAuth/Amplify+Cognito, SWCamera+Vision, SWPaywall/StoreKit 2, SWChat with VolcEngine ASR, SWTikTokTracking) — exposed via `listRecipes`, `getRecipe`, `searchRecipes` for iOS 18+/Swift 5 apps.
Local-first code intelligence MCP server that pairs the research-backed cAST chunking algorithm with multi-hop semantic search and regex — indexes 30 languages (Python, JS/TS, JSX/TSX, Java, Kotlin, C/C++/C#, Go, Rust, Swift, Vue, Svelte, Zig and more plus JSON/YAML/TOML/HCL/Markdown/PDF), supports VoyageAI/OpenAI/Ollama embeddings with Claude Code CLI or Codex CLI for code research, and integrates with Claude/VS Code/Cursor/Windsurf/Zed via MCP.
Provides deep automated codebase intelligence through hybrid semantic and BM25 search, AST-aware code chunking, polyglot dependency graphs, and searchable infrastructure artifacts — all zero-config, fully local, and tested on repositories with over 40 million lines of code.
Aggregates, filters, and distributes news articles from multiple sources using advanced machine learning and customizable content pipelines.
ArcadeDB is a high-performance multi-model DBMS (graph, document, key/value, search, time-series, vector, geospatial) with Java 17/21 LLJ engine, SQL/Cypher/Gremlin/GraphQL/MongoDB query languages, 70+ built-in graph algorithms, parallel query execution, and a built-in Model Context Protocol (MCP) server so AI assistants and LLMs can query the database directly.
Give Claude, Cursor, or any MCP client a real email address — free, no signup. The `@keyid/agent-kit` MCP server exposes 27 tools for autonomous email: identity/auth (`keyid_provision`, `keyid_get_email`), inbox & send (`keyid_get_inbox`, `keyid_send`, `keyid_reply`, `keyid_forward`), threads/drafts, auto-reply/signature/forwarding settings, contacts, webhooks, and usage metrics. Uses Ed25519 keypairs and MCP JSON-RPC over stdio (protocol 2024-11-05).
An agent-native travel search MCP server that finds the cheapest flights and hotels by querying 75+ airline connectors and enterprise GDS sources (Amadeus, Duffel, Sabre, Travelport) in parallel, returning results in about 5 seconds with zero markup.
MCP server that validates startup or side-project ideas by querying real data from GitHub, Hacker News, npm, PyPI, and Product Hunt in parallel and returning a 0-100 `reality_signal` with top competitors and pivot hints via the `idea_check` tool. Uses Claude Haiku 4.5 for multilingual query generation with a dictionary fallback.
MCP server for VocoType, a privacy-focused offline voice input tool that converts speech to text locally without uploading any data, supporting MCP voice-to-text and AI text optimization.
Turns Claude Code into a design assistant with local ComfyUI image generation, multiple API provider integrations, a library of 1,300+ professional prompt templates, and multi-direction parallel output for rapid visual creation.
Manages Twitter API interactions, authentication, and data retrieval for distributed social media applications and analytics platforms
MindOS is a local-first, human-AI collaborative knowledge base that provides a shared memory layer for all AI agents, storing notes, workflows, and personal context as plain-text files that are both human-readable and directly executable by agents.
MCP server for JavaScript reverse engineering. Gives AI clients (Claude, Codex, Cursor) unified browser-debugging tools for script discovery, breakpoints, runtime hooks, network/request-chain analysis, obfuscation recovery, and risk assessment — replacing the usual DevTools + scripts + packet-capture juggling.
Fast web scraper MCP server for AI agents — sub-millisecond extraction with 67% fewer tokens and zero browser overhead. Installed via `npx create-webclaw`.
Memorix is a persistent memory MCP server for AI coding agents that stores decisions, debugging history, and project knowledge across sessions and IDEs, allowing multiple agents to share context without losing information between sessions.
Graph-powered code intelligence engine that indexes codebases into a knowledge graph and exposes it via MCP tools. Provides blast radius analysis, dead code detection, execution flow tracing, and git coupling analysis for AI agents.
Open-source agent firewall that intercepts outbound AI-agent HTTP requests to block data exfiltration — enforcing URL allow/block rules, DLP against secrets like `ANTHROPIC_API_KEY`, and configurable safety presets. Ships as a single Go binary with zero runtime dependencies and integrates with Claude Code, OpenAI Agents SDK, Google ADK, AutoGen, CrewAI, LangGraph, and Cursor.