How to Track On-Chain Usage of AI Agent Tokens with Dune Analytics

 / 
2

The core of tracking AI Agent tokens isn't about price, but about actual on-chain invocations. With Dune, there are two approaches: directly search for dashboards built by the community to see data, or use Dune MCP to let AI write SQL queries for smart contract call volumes of specific protocols. The latter is currently the most accurate way to define "on-chain usage" — because the essence of an AI Agent is a machine calling smart contracts, not a person transferring tokens with a wallet.

Search Community Dashboards: View On-Chain Data for AI Agent Tokens

What to do: Search directly on Dune for existing dashboards of AI Agent tokens (such as ai16z, VIRTUAL, etc.) and review data panels that the community has already organized.

How to do it:

  1. Open the Dune Analytics website (dune.com) and go to the search bar.

  2. Enter the AI Agent token name, for example "ai16z". Dune's community-driven nature means most popular Agent projects already have dashboards created by users.

  3. Find the top-ranked dashboards and check the content they contain:

    • Token holdings: the composition of tokens held in the project's wallet

    • DEX trading volume: daily trading volume of the token on decentralized exchanges

    • Holder address distribution: concentration of holdings among large holders

What counts as done: Find 1–2 dashboards with relatively comprehensive coverage so that you can view basic usage data for that Agent token.

The advantage of community dashboards is that they are ready to use immediately and require no SQL. The downside is that the update frequency and the metrics covered depend on the creator and may not include granular indicators such as "smart contract call volume".

Query On-Chain Call Volume Using Dune MCP with Natural Language

What to do: Use the Dune MCP server to query smart contract call volumes related to AI Agent tokens in natural language — this is the most direct way to track "machine usage".

How to do it:

  1. Set up Dune MCP: Add the Dune MCP server in an AI client that supports MCP (such as Claude Code, Cursor, OpenCode). Dune MCP supports both OAuth 2.0 and API Key authentication; OAuth is recommended for browser-based environment configuration.

  2. Query Agent protocol call volume: The payments.agentic_payments table in Dune specifically records machine-to-machine on-chain payment activity, covering x402 and MPP protocols. You can say to the AI assistant in natural language: "Query daily transaction count and total amount for the x402 protocol over the past 7 days, displayed by chain for Base, Polygon, and Solana"

  3. View call details: Dune MCP exposes 12 core tools, including searchTables (discover relevant tables), executeQueryById (execute a query), and getExecutionResults (retrieve results). The MCP automatically completes the whole workflow: table discovery → SQL writing → execution → returning results.

What counts as done: Successfully query on-chain call data for an AI Agent protocol (such as x402) or a specific Agent token through MCP.

x402 and MPP are the underlying protocols for AI Agent automatic payments. x402 settles on-chain via EIP-3009 or Permit2; MPP's Tempo rails support both one-time payments and long-running sessions. These call volumes directly reflect the "usage" of an AI Agent.

Check Direct Calls to an Agent Token's Smart Contract

What to do: Query the number of calls to specific functions of an AI Agent token contract to determine whether it is actually being used.

How to do it:

  1. Confirm the token contract address: First find the Agent token's contract address on Etherscan or Solscan.

  2. Locate relevant tables in Dune: Use the searchTables tool to find the decoded event table associated with that contract. If the Agent has specific "invocation" functions (such as executing a task, triggering an action), the corresponding event logs will more directly reflect usage.

  3. Query call volume: Generate SQL to count the number of calls to the specific function. It is enough to count the call data for the last 30 days.

What counts as done: Obtain the function call count and the number of unique callers for that Agent token over the past 30 days.

Common reasons for failure:

  • Mixing up "token transfers" and "contract calls": Transfer volume between holding wallets reflects trading speculation, not Agent usage. True "on-chain usage" refers to the number of times an AI Agent calls a smart contract to perform a task (such as payment or data request).

  • No ready-made dashboard found on Dune: Newer Agent tokens may not yet have a community dashboard, so you need to query on your own using MCP.

Risk reminder:

  • Dune API is usage-based: MCP itself does not incur extra fees, but queries consume Dune API credits. Use the getUsage tool to monitor consumption.

  • Agent token ≠ protocol usage: The price swings of many AI Agent tokens have nothing to do with their actual on-chain call volume. When querying data on Dune, focus on "smart contract call count" rather than "transfer volume". The latter reflects trading heat, not product usage.

How to confirm you have mastered the tracking method?

Search on Dune for any AI Agent token you follow, find a community dashboard, and check its holdings and trading volume data. If you want to dig deeper into "call volume", set up Dune MCP as described above, run a query and confirm that it returns data. From there, you can continuously track the weekly call trends of the Agent protocols you monitor, observe how call volume changes relate to token price movements, and judge whether market pricing matches real usage.