Skip to main content
Your agent finds a tool in the marketplace, checks its inputs and price, and uses it through AgentMuxer.
1

Find a tool

Your agent searches the marketplace for a tool that fits your task. Search is free.
2

Check the tool

Your agent checks the tool’s inputs, price, and available test results before using it.
3

Use it within your budget

Your agent supplies the inputs and a maximum price. AgentMuxer checks your balance, spending limits, and any required confirmation, then sends the request to the provider.
4

Review the result

Your agent receives the result and cost, and can report whether the tool helped.

​
Who does what?

Your agent chooses the tool. Your model still runs with your existing provider. For tool names and parameters, see the MCP reference.

​
What the listing badges mean

Badges show what has been checked. They help you choose a tool, but don’t guarantee the result.
The listing includes task-evaluation evidence. Read the evidence to understand what was tested and whether it matches your use case.
The tool’s published input definition has been retrieved and checked. This does not mean a paid call or its results have been tested.
The listing was discovered in a public index and verification is pending.

​
MCP and payments

AgentMuxer exposes its tools through MCP, which lets agent clients and frameworks discover and call them. The TypeScript SDK adapts that connection to your framework. AgentMuxer handles provider payments, including MPP and x402. You don’t need to integrate either protocol yourself.

Build an agent

Add AgentMuxer to an OpenAI agent.

Connect your agent

Set up an agent you already use.