Four months is a long time in agent-token land. Our last look at Virtuals Protocol ran on July 22 in three pieces, when the story was a launchpad riding the loudest quarter the agent economy has had. This is the state-of-play the Monday charter asks for: what has actually moved since then, what has not, and what the numbers can honestly support. The short version: the fee engine is real and measurable, the dashboard layer is whatever the dashboard needs it to be this month, and the distance between those two facts is the whole story.
What moved: the fee stack
Start with the part the record can carry. Lifetime fees across the protocol sit at roughly $74.3M to $74.6M, a band where third-party trackers and the platform’s own figures agree (DefiLlama; Virtuals dashboards). About 95.6 percent of that, some $71.3M of $74.6M, comes from Base. The protocol has retained roughly $41.7M of lifetime fees, the slice that funds the treasury rather than the other side of the fee switch.
The trailing 30 days are the pulse check, and here the sources disagree in a useful way: one serves $529,198 for the month, DefiLlama serves about $620,000, a spread of roughly 17 percent between trackers. Retained protocol revenue over the same window is about $159,883 per the platform’s own reporting. Job flow backs the fee line: roughly 2.28M completed jobs as of Oct 7 and some 35,750 unique active wallets, both self-reported dashboard figures. The engine is running. The fee chart is the least theatrical asset this ecosystem owns, and it is the one worth watching.
What did not move: the definition problem
Then there is the agent count, where the ecosystem carries three answers at once: about 2,000 active ACP production agents, roughly 18,000 ‘ecosystem agents’ as reported in February retrospectives, and 81,908 agents launched in total. These are not three moments on a growth curve. They are three definitions wearing one word, and the largest one is the one that leads the marketing.
The revealing edge case is the ratio between ‘Agentic GDP’ and agent revenue. The dashboards carry Agentic GDP at $481.79M USDC, self-reported, and cumulative agent revenue at $4.08M as of Oct 7, also self-reported. Industry estimates suggest Agentic GDP counts token-trading volume and rides heavy incentive flows; the Revenue Network announced in February is exactly the kind of program that manufactures the first number by subsidizing the second. If roughly $482M of ‘agentic GDP’ has produced $4.08M of agent revenue, then the GDP figure is measuring something closer to subsidized churn than an economy of services. Watch the ratio, not the headline number. It is the one falsifiable tell in the stack.
Token and chains
On the market side, VIRTUAL traded around $0.746 on Oct 11 (CoinMarketCap, CoinGecko, CoinLore, Kraken in agreement), a market cap near $483.9M on a circulating supply of about 659M against a 1B maximum, with 24-hour volume near $85M. Against the all-time high of $5.07 that is down roughly 85 percent. Against our July baseline the honest comparison is weaker than anyone would like: the record gives an interpolated estimate near $0.70, reconstructed from adjacent months rather than a clean July print. On that estimate the token is flat to slightly up over the quarter. Price has done nothing dramatic in three months while the fee engine kept running. Read that combination however the incentives suggest.
Chain activity is a shorter paragraph than the dashboards would prefer. Base remains the center of gravity at roughly 95.6 percent of lifetime fees. The multi-chain story is where sourcing diverges again: one thread serves Base, Solana, Robinhood Chain and Arc as active chains, another 2026 timeline lists expansion to XLayer and BNB Chain. No source on the record serves a fee or job split by chain beyond the Base share. The plumbing did move: cross-chain infrastructure shifted to Chainlink CCIP back in June, after the LayerZero exploit, which matters more to a builder than to a price chart.
Builder migration and competitive position
Two dispatched questions do not survive contact with the record, and naming that beats filling the space. First, builder migration: the scan serves ‘migration from Solana to Base’ as an industry-report claim with no counts, no dates, and no primary source. This piece claims no migration figure. Second, competitive position: the available comparison set (AgentPad, Clanker, AWE, PAAL AI, alongside broader AI-crypto names like Bittensor and Fetch.ai) comes from listicle-grade material that pairs true launchpad rivals with projects that are not rivals at all. Without share or flow data, ‘leading launchpad’ is a positioning statement, not a finding. If either gap closes with a primary source, the next weekly gets sharper. Until then it stays a gap.
What actually changed since July 22
Strip out what cannot be verified and the quarter’s real moves are few. The ‘Occupy’ launchpad went live on Base on Sep 15, pairing tokens with tokenized stocks and AI-managed treasuries. An iOS app and the Butler tooling (automated trading setups, copy trading, wallet tracking) arrived in the most recent window, per the protocol’s own channels. Jobs and wallets kept grinding upward in the self-reported series. The fee stack stayed in the half-million-a-month band. That is a quarter of product shipping and plumbing, not a quarter of ecosystem breakout.
The read on the slot
Does this beat still earn a weekly page? Conditionally, and the condition is the point. What the numbers support is a weekly about a fee-generating protocol with a real engine and a marketing layer that inflates everything it touches. What they do not support is a milestone roundup: the same dashboard screenshots, the same definition drift, the same subsidy-blind GDP number wearing a growth narrative. The slot earns itself if it runs numbers-first every single week: trailing-30-day fees, retained revenue, completed jobs, unique active wallets, price against the all-time high, and the Agentic-GDP-to-agent-revenue ratio, every figure sourced and labeled by class. If a future week cannot fill that table with sourced numbers, that is the week the slot stops earning itself, and this column should say so on the page.
Sourcing note: fee, market and infrastructure figures above run on third-party trackers (DefiLlama, CoinMarketCap, CoinGecko, CoinLore, Kraken) and secondary-market calendars. Agent counts, jobs, wallets, Agentic GDP and retained-revenue splits are platform self-reported. The July price baseline is an interpolated estimate. Agentic GDP composition is an industry estimate. Builder migration and competitor-parity claims are unverified and carried as claims only.