FunASR is an industrial-grade speech recognition toolkit that runs 170x faster than realtime, supports 50+ languages, and includes speaker diarization, emotion detection, and streaming. It exposes an OpenAI-compatible API and can be run as an MCP server for integration with Claude and Cursor.
BlockRun MCP is a pay-per-call server that gives Claude (and other MCP clients) live data from markets, crypto, search/research APIs, and Twitter/X. You fund a local wallet with a few dollars of USDC (on Base or Solana) and each API call deducts a sub-cent fee. No API keys, no subscriptions, no dashboard.
Cinderwright is an MCP server and universal payment router that lets AI agents pay for and access services using USDC (x402 protocol) or Bitcoin Lightning (L402). It also acts as a discovery hub claiming over 2,800 services, providing a proxy API so agents don't need to manage wallets or payment protocols directly.
MCP server that lets you browse OrcaRouter's model catalog and run chat completions through various LLMs (Anthropic, OpenAI, DeepSeek, Gemini, etc.) with auto-routing and fallback chains, directly from MCP clients like Claude Desktop, Cursor, and Windsurf.
A registry service that acts like an app store for MCP servers, letting clients discover and connect to available servers. Built in Go and distributed via Docker, it provides a community-driven catalog with an API for publishing and querying MCP server metadata.
An MCP server that acts as a Provider Abstraction Layer, letting you route requests across multiple LLM providers (OpenAI, Anthropic, Google, Grok, Azure, Ollama, etc.) from a single context. Its standout feature, clink, can spawn external AI CLIs as isolated subagents—like having Claude Code launch a Codex subagent for a fresh-context code review—while preserving conversation continuity between tools.
MCP server that lets you use DeepSeek's chat and reasoning models from any MCP-compatible client. Supports multi-turn conversations, function calling, thinking mode, and cost tracking. Can run locally via npx or connect to a remote hosted endpoint.
Headroom is a context compression layer that sits between AI agents and LLMs. It exposes a Python/TypeScript library, a local proxy, and an MCP server to shrink tool outputs, logs, files, RAG chunks, and conversation history before they reach the model, claiming 60-95% token reduction while preserving answer quality.
MaxKB is an open-source platform for building enterprise-grade agents and knowledge base chatbots. It uses a RAG pipeline to ingest documents, supports multiple LLMs (private and public), and offers MCP tool-use capabilities through a workflow engine.
MCP server that lets AI agents discover each other's capabilities, request tasks, and settle payments in SOL—all peer-to-peer over Nostr relays with no central platform.
AgentForge is a marketplace and unified API gateway that lets you access over 300 AI agents through a single API key. It also provides an MCP server that exposes these agents as tools for Claude Desktop, Cursor, and other MCP clients.
Roundtable AI is a local MCP server that lets you orchestrate multiple AI coding assistants (like OpenAI Codex, Claude Code, Cursor, and Gemini CLI) in parallel from a single prompt in your IDE. It discovers your installed CLI tools automatically, delegates sub-tasks to specialized models, and synthesizes results—removing the need to copy-paste between different AI tools.