Good breakdown of practical MCP tools for agent perception. One angle worth flagging: each Apify call has a real cost, and agents making autonomous decisions about which URLs to screenshot or audit can burn through a budget fast without realizing it. Pairing these tools with a programmable wallet that enforces per-tool spending caps gives the agent autonomy to operate within limits without needing a human to approve every screenshot. The same applies to the legitimacy checker, great for gating risky actions, but the check itself should also be metered.