Mercator lets agents discover, pay for, and combine use outside tools, improving their performance on domain specific tasks along the way.
Tempo has launched Mercator, a tool router that lets AI agents find, pay for, and use outside services on demand instead of requiring developers to configure every tool in advance.
What’s the Scoop?
-
Tools on demand: With Mercator, an agent simply describes what it needs and its budget. Mercator then searches available services, ranks them by fit, reliability, and cost, and executes the selected workflow.
-
The agent multi-tool: The idea should be familiar. MPP already launched with a directory, while we’ve seen a series of similar “agent multi-tools,” including AgentCash and Pay.sh, that give agents one place to discover and pay for outside APIs without separate accounts or keys. Mercator pushes this model further: rather than just helping an agent find and access a tool, it can rank services and combine multiple paid tools into a single workflow.
-
One payment layer: Mercator connects agents to services using open payment standards MPP and x402, while users pay Mercator itself primarily with MACH, a USD-denominated Tempo credit purchasable through Apple Pay. It also supports USDC.e and pathUSD payment routes, handling payments to downstream services behind the scenes.
-
Making smaller models better: Tempo says Mercator improved model performance across all four benchmarks it tested, with the largest gains coming from cheaper models. GPT-5.6 Luna at high effort, for example, improved from 66.8% to 70.3% across the benchmark suite with Mercator, putting it within 2.3 points of GPT-5.6 Sol at medium effort without it.
Be the first to comment