mcpdirectory.dev
9,129 servers · updated 23 Sept

Proxyllm MCP server

community Active last commit 2 mo
{
  "mcpServers": {
    "proxyllm": {
      "command": "npx",
      "args": [
        "-y",
        "proxyllm-mcp@latest"
      ],
      "env": {
        "PROXYLLM_ACCOUNT_TOKEN": "<proxyllm_account_token>",
        "PROXYLLM_ROUTING_KEY": "<proxyllm_routing_key>"
      }
    }
  }
}
file .mcp.json (in your project root)install npx -y proxyllm-mcpversion 1.0.2

What it does

MCP server for ProxyLLM, an OpenAI-compatible LLM gateway. It exposes model catalogs, cost calculation, routing key management, and an autonomous account signup flow. For developers integrating with ProxyLLM or evaluating its pricing and available models.

  • Calculate monthly savings by comparing your current OpenAI bill against ProxyLLM subscription tiers.
  • List available models from the ProxyLLM gateway and check what a routing key can access.
  • Complete account signup and provisioning end-to-end: send verification codes, exchange them for account tokens, and create routing keys.

Tools this server exposes

ToolDoes
proxyllm_savings_calculatorCalculate subscription tier recommendation and monthly savings from an OpenAI bill amount.
proxyllm_list_modelsFetch the OpenAI-shaped model list available from the ProxyLLM gateway.
proxyllm_introspect_keyCheck what a routing key can access: lanes, models, budget, and subscription backing.
proxyllm_signupSend a 6-digit verification code to an email address for account creation.
proxyllm_verify_signupExchange a verification code for an account token.
proxyllm_accountCheck account state, payment status, and checkout link if payment is required.
proxyllm_create_routing_keyCreate a routing key with optional provider chain configuration.
proxyllm_list_routing_keysList routing keys and optionally the lane ids available for wiring.
proxyllm_usageGet a 30-day usage summary for the account.

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.