LLM Broker

net.llm-brokerv1.0.0Updated Oct 2, 2026

One key, every model: measured scores and live prices, routed per request to the cheapest fit.

VerifiedStreamable HTTPWeb executableAI & ML

Overview

AI-generated overview

A remote MCP endpoint that routes chat requests across many LLM providers using one API key, picking the cheapest model that fits.

What it does
LLM Broker is a hosted MCP server that exposes a chat tool and a model-listing tool. It routes each chat request to a model chosen from measured scores and live prices, aiming for the cheapest option that fits the request. Listing available models requires no key; sending chat requests does.
When to use it
Useful when an assistant needs access to many models through a single credential instead of separate provider accounts, or when cost-aware routing between models matters. Less useful if you already have direct provider access or need a specific fixed model.
Requirements
A remote streamable HTTP endpoint; nothing is installed locally. Chat requests need an Authorization header with a bearer token of the form ast_sk_…; listing models needs no key. Network access to the provider's API endpoint is required.
Before you install
The chat tool sends your prompts to a third-party broker, which forwards them to whichever model it selects, so prompt content leaves your control. The Authorization bearer token is a billable credential; treat it as a secret and expect usage-based charges. Routing decisions are made by the service, so you may not know which model handled a request.

Installation

In SourceWeft

  1. Open LLM Broker in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Other MCP clients

Add this to your client's mcpServers config.

{
  "mcpServers": {
    "broker": {
      "type": "http",
      "url": "https://api.llm-broker.net/api/v1/broker/mcp"
    }
  }
}

Tools

0
Tool metadata has not been indexed yet.

Version history

1
  1. v1.0.0LatestOct 2, 2026