Tempo has introduced Mercator, a new tool router that enables AI agents to find, pay for, and utilize third-party services as needed, rather than requiring developers to pre-configure each tool. With Mercator, the agent simply describes its needs and budget. In response, the system searches for available services, ranks them based on suitability, reliability, and cost, and then executes the selected workflow. Mercator expands on the previously introduced 'multitasking tools for agents,' such as AgentCash and Pay.sh, which provide agents with a unified interface for using third-party APIs without the need to create separate accounts.
Mercator allows multiple paid tools to be combined into a single workflow. It connects agents to services using open payment standards MPP and x402, with users primarily paying for Mercator using MACH, a credit tied to the US dollar and acquired through Apple Pay. Additionally, payment routes USDC.e and pathUSD are supported, processing payments in the background.
According to the company, Mercator has improved AI model performance across all four tested benchmarks. The greatest gains were noted when using cheaper models. For instance, in high-effort mode, GPT-5.6 Luna showed an increase from 66.8% to 70.3% on benchmarks, which is only 2.3 points lower than GPT-5.6 Sol at average effort without using Mercator.
Related articles
