The measurement layer the agent economy is missing
AI agents are already buying things, signing contracts, and moving money. The infrastructure they run on is being built fast, in the open, by a handful of large platforms. What is missing is someone independent measuring whether any of it is safe, adopted, or real. That is the gap Major Labs builds into.
How we work
We measure what others assert, we do it in the open, and we ship the tools alongside the research.
The ecosystem is described in marketing numbers. We go and count: 38,000 MCP servers advertised, about 1,200 genuinely usable, roughly a third of the most-used carrying a risky pattern. Firsthand, every time.
Every number is checkable. The scanner is a couple hundred lines of readable Python, the dataset is public, and we would rather you argue with the method than take it on faith. Credibility is the product.
We never connect to, run, or probe anyone's system. We read public source and public signals. A finding is a pattern, not an exploit, and crossing that line once would cost the trust everything else is built on.
Per-repo problems go privately to the people who can fix them, on a clock. We publish the aggregate and we name the good actors who fix things. We do not publish target lists.
The research ships next to the tools it implies: open-source primitives that govern what agents do, and a scanner anyone can run in their own CI. The studio practices what it measures.
What we ship
- Reports. State of MCP, State of AEO, and State of Agent Memory — firsthand findings with one sticky, checkable number each.
- Open data. the live security scoreboard and the versioned dataset — re-swept and updated, so the question becomes the trend, not a single snapshot.
- Open-source tools. Agent-safety primitives (IdentityKit, MandateKit, BudgetGuard, WitnessKit, RememberKit) and the scanner behind the scoreboard, all on GitHub.
Who's behind it
Major Labs is built by Charlie Major. Fifteen years in payments across product and engineering, building the infrastructure most of the world uses without ever knowing it is there, now turned toward the agent economy. He reads a pull request and a P&L with equal fluency, and he ships the tools, not just the takes. He also writes Major Matters, the publication on payments, AI, and commerce. Major Labs is where the research and the code live; Major Matters is where the analysis lands. Both are independent, and the opinions are his own.
Talk to us
Building on this, fixing a finding, or want the method picked apart? hello@majormatters.co.