Side-by-side comparison of two Model Context Protocol servers — install paths, tools, quality signals, and directory engagement so you can pick the right one for Claude, Cursor, and other MCP clients.
Connect your agent to any HTTP API on the fly — discovers + maps any REST API once, then fetches typed data deterministically (no per-call LLM). Self-hosted MCP server (uvx --from 'liquid-api[mcp]' liquid-mcp); works with OpenAI/Gemini/Anthropic/local or any provider via LiteLLM. Open source (AGPL).
Governance proxy for MCP: pair once and route your MCP servers through policy gates (allow / deny / human-in-the-loop approval), DLP redaction before the model, and cost caps, with every tool call recorded to a hash-chained audit log exportable as an Ed25519-signed, offline-verifiable evidence pack. Single-file binary daemon (macOS / Linux / Windows); downloads and install docs in the linked repo.