Agent Charts for agents
Your agent can ask the ranking itself. Connect it to our MCP server for free, or let it pay $0.01 a query over x402, and it gets the numbers this site shows: which paid tools AI agents really pay for, with wash trading filtered out, in the 30 days to 9 Nov 2026.
Connect in a minute
In Claude Code:
claude mcp add --transport http agentcharts https://agents.agentcharts.xyz/mcpIn Cursor or another MCP client, add a remote server over streamable HTTP at https://agents.agentcharts.xyz/mcp. There's no key and no sign-up. Then ask in plain words, for example "find me the best inference provider for my agent" or "which web search API do agents pay for most?"
What the agent gets
Asked for an inference provider, the agent calls best_tools for Inference & gateways and gets the category's ranking. The top three, in the 30 days to 9 Nov 2026:
| Rank | Seller | Spent in 30 days | Buyers | Paid calls | Typical price per call |
|---|---|---|---|---|---|
| 1 | dtelecom.org | $36,721 | 30 | 16,756 | $1.00 |
| 2 | blockrun.ai | $5,853 | 653 | 781,830 | $0.001 |
| 3 | venice.ai | $1,201 | 42 | 988 | $0.01 |
A good agent reads the rows the way a person would: the top seller by money may have a few big buyers, and one further down may have many more at a lower price. Pick a paid tool explains each column.
Free or paid
| Door | For | Price | Address |
|---|---|---|---|
| MCP server | Agents a person sets up, such as Claude Code | Free | https://agents.agentcharts.xyz/mcp |
| x402 endpoint | Agents that find and pay for services on their own, with no account | $0.01 a query, in USDC on Base | https://agents.agentcharts.xyz/v1/best-tools?category=<slug> and https://agents.agentcharts.xyz/v1/seller/<domain> |
Both answer with the same three things: categories lists the categories, best_tools ranks one category (optionally only tools under a price per call), and seller gives one tool's numbers. Machines can read the details in llms.txt and agent.json, and people on the service's own page.
Why these numbers
Every number comes from payments on public blockchains, over x402 and MPP. Before anything is ranked, the filter takes out sellers paying themselves, sellers with fewer than 10 buyers and money that goes back, and flagged sellers are reviewed every Monday. Nobody can pay for a place. How the ranking works has every rule.
Questions
What is the Agent Charts MCP server?
A free MCP server at https://agents.agentcharts.xyz/mcp. It gives an agent three tools, categories, best_tools and seller, that answer from the same data as this site: which paid tools AI agents really pay for, by money, buyers and price, over the last 30 days.
Is it free?
The MCP server is free, with no key and no sign-up. The x402 endpoint charges $0.01 a query in USDC on Base, paid by the agent with each request. The category list is free on both.
Which agents and clients can use it?
Any MCP client that can add a remote server over streamable HTTP, such as Claude Code or Cursor. Agents that pay for services on their own can use the x402 endpoint with any x402 client.
How fresh is the data?
The numbers update daily and cover the last 30 days; the data now runs to 9 Nov 2026. Sellers flagged by the filter are reviewed every Monday.
Can a seller pay for a better place?
No. Places come only from what agents paid on-chain, after the filter. The money paid to our own endpoint is shown on the removals page and never ranked.
What can't the ranking tell my agent?
It counts what agents paid, not how good a tool is. It doesn't test quality, speed or uptime, and it leaves out free tools and card payments. Try a tool on a few calls before you build on it.