TokenEase MCP Server

OpenAI-compatible chat completions gateway for Chinese LLMs (Kimi K3, GLM-4-Flash, Hunyuan) over SSE. One Bearer token, OpenAI-style base_url swap, 1M free trial tokens.

View on tokenease.io

Hosted MCP Server

npx add-mcp 'https://tokenease.io/mcp/sse'

Installs into Claude Code, Codex, Cursor and more

Documentation

Live right now: 4 of 17 models — Kimi K3, Kimi K2.7 Code, Kimi K2.7 Code HighSpeed and GLM-4 Flash. Every other model name on this page (DeepSeek, Qwen, Doubao, GLM-5.x, Tencent Hunyuan) is paused and returns 503 model_unavailable. Always read GET /v1/models for the live list.

TokenEase MCP Server

Connect Claude Desktop, Cursor, Continue.dev and any MCP client to Kimi K3, Kimi K2.7 Code, Kimi K2.7 Code HighSpeed, and GLM-4 Flash. One URL, zero config files.

🔌 SSE Transport

Real-time Server-Sent Events for dynamic tool discovery and invocation

🤖 Models Live Now

Kimi K3, Kimi K2.7 Code, Kimi K2.7 Code HighSpeed, GLM-4 Flash

⚡ One URL Setup

Paste https://tokenease.io/mcp/sse into your MCP client — done

💰 Pay Per Token

Use your existing TokenEase API key. No separate billing

Quick Setup

Get Your API Key

Claude Desktop

Open Claude Desktop → Settings → Developer → Edit Config. Add:

{
  "mcpServers": {
    "tokenease": {
      "url": "https://tokenease.io/mcp/sse",
      "headers": {
        "Authorization": "Bearer YOUR_TOKENEASE_API_KEY"
      }
    }
  }
}

Cursor

Open Cursor → Settings → MCP → Add new MCP server:

Name: TokenEase
Type: SSE
URL: https://tokenease.io/mcp/sse
Headers: Authorization: Bearer YOUR_TOKENEASE_API_KEY

Continue.dev (VS Code)

Add to your config.json:

{
  "experimental": {
    "modelContextProtocolServers": [
      {
        "url": "https://tokenease.io/mcp/sse",
        "headers": {
          "Authorization": "Bearer YOUR_TOKENEASE_API_KEY"
        }
      }
    ]
  }
}

Available Tools

Tool	Description
`tokenease_chat`	Send chat completions to any supported model
`tokenease_list_models`	List all available models with pricing and capabilities

Tip: Once connected, ask Claude/Cursor "Use tokenease_chat with kimi-k3 to explain this code" and it will automatically route through TokenEase.

Supported Models

kimi-k3

Kimi K3 — 256K context, MMLU-Pro #1. $0.28/M input

kimi-k3

Kimi K3 — 256K context, MoE architecture. $0.28/M input

glm-4-flash

GLM-4 Flash — Fastest & cheapest. $0.10/M input

kimi-k2.7-code

Kimi K2.7 Code — coding specialist. $1.20/M input

kimi-k2.7-code-highspeed

Kimi K2.7 Code HighSpeed — fastest coding tier. $2.50/M input

glm-4-flash

GLM-4 Flash — cheapest tier. $0.10/M input

New: Use DeepSeek V4 with Claude Desktop

Step-by-step guide to connect Claude Desktop to DeepSeek, Kimi K3, and 15+ models via MCP.

View Setup Guide →