The hype cycle around autonomous AI agents is colliding with the mundane reality of broken API keys. Agent Earnings, a niche directory tracking revenue for agent-run businesses, disclosed on November 13, 2026, that 6 of its 106 'verified' entries rely on data from dashboards that had stopped syncing. These figures, totaling $8,836 in monthly recurring revenue (MRR), were pulled from TrustMRR pages where the underlying founder API keys had expired. While this represents only 1.3% of the directory's $666,935 combined MRR, it exposes a critical flaw in how we validate 'autonomous' income: verification often means a script read a Stripe dashboard, not that the agent is currently generating cash.

The Illusion of 'Verified' Status

The term 'verified' in the AI agent space has been dangerously loosely defined. According to the Agent Earnings post, 105 of the 106 entries derive from Agent Money List snapshots, which pull MRR via TrustMRR. The remaining entry, Moltcorp, was read directly from TrustMRR. 'Verified' simply confirms that a service read the number from a payment provider like Stripe, Polar, or Whop, rather than relying on a founder's self-reported spreadsheet. It does not imply an audit of costs, profit margins, or the degree of autonomy. Crucially, this verification status is only valid while the API key remains active. When keys expire, as they did for Client One, BillionVerify, PrimeClaws, and Access Labs, the data freezes. For four of these six cases, the data was already months old when the October 16, 2026 snapshot was taken.

Median Skew and Data Integrity

The presence of stale data distorts the perceived health of the AI agent economy. The directory’s median MRR dropped from $4,380 on November 10 to $2,806 by November 13, largely due to the addition of 30 lower-earning businesses from the bottom of the Agent Money List rankings. However, the six stale entries are all below the current median. Excluding them would raise the median to $3,416, painting a slightly rosier picture of agent profitability. The directory maintains these entries with explicit caveats rather than silently removing them, acknowledging that a real payment record exists, even if it isn't current. Access Labs, for instance, shows conflicting descriptions between Agent Money List (AI-adoption consulting) and its TrustMRR profile (NFT membership and web3 development), highlighting the chaotic metadata surrounding these projects.

The Reality Check: Zeroes and Roasts

Beyond the stale figures, the directory includes self-reported runs that generated zero revenue, challenging the narrative that every autonomous agent is a profitable enterprise. Idle Hands Labs, an AI building a startup in public with a goal of $1,000 in revenue by November 30, 2026, reported on day 25 that it still had $0 revenue. Rather than hiding the failure, the agent is giving away its '5 USDC deep wallet roast' for free to the first 10 respondents. The agent’s rationale is stark: 'a paid product with zero users teaches me nothing.' This transparency contrasts sharply with the opaque, often inflated MRR claims seen elsewhere in the ecosystem, where 'verified' might just mean 'we scraped a dead dashboard.'

Key Takeaways

  • 'Verified' MRR in AI agent directories often means a snapshot of a payment provider's dashboard, not a live audit of autonomous activity.
  • Expired API keys can freeze revenue data for months, creating a false sense of current performance for agent-run businesses.
  • The median MRR of the AI agent economy is highly sensitive to which businesses are included and whether stale data is filtered.
  • Some agents openly report zero revenue, prioritizing user feedback over inflated financial metrics, offering a more honest view of early-stage autonomy.

The Bottom Line

Stop treating scraped MRR as proof of autonomous intelligence. If the API key is dead, the agent is just a ghost in the machine, billing nobody.