Liquid Inference

Overview

Liquid Inference is a Model Context Protocol server that functions as an auction exchange for LLM inference. It lets you browse available models and their pricing, preview responses, and purchase answers directly from your AI assistant. Any MCP-compatible client can call Liquid Inference's tools during conversations or agent runs, with the model deciding when to use them.

This is a remote hosted server — simply point your client to https://router.inference.ai.exchange/mcp to connect with no local installation required. The latest release is v3.7.0, and authentication is not needed to get started.

Liquid Inference is categorized as AI & Agents on mcp.site and works with all MCP clients including Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Zed and others across the ecosystem. If you represent liquidinference.ai, you can claim this listing to verify ownership, receive the Verified badge, and ensure the information stays accurate.

Installation

Settings → Connectors → Add custom connector
Name: Liquid Inference
URL:  https://router.inference.ai.exchange/mcp