AI agents are moving from novelty to infrastructure. As they gain the ability to access data, spend money, and act with delegated authority, a basic question keeps going unanswered in public: who is actually behind this, and what has it been authorized to do?
Operator Record is an independent directory built to answer that question with structure and sources, not marketing claims. We are not a certification body and we do not rate agents as "safe." We document what is known, what is claimed, and what remains unverified — clearly labeled at every level.
Five categories today: personal assistants, enterprise agents, developer frameworks, commerce and payment agents, and agent identity and security infrastructure. Each listing follows the same schema, whether it's a household-name product or a small open-source project.
Operator Record sits alongside Proof Human, which asks the parallel question about people: who is the human, and how is identity or uniqueness established? Together the shared theme isn't just "proof" — it's accountability for digital actors, whether human or machine.
Every listing in the directory has been checked against primary sources — official documentation, public repositories, audit reports and regulatory filings — following the process described in our methodology. Each profile carries an evidence level showing how well its claims are supported, and a date recording when it was last reviewed.
This is an actively maintained directory rather than a finished reference work. The agent landscape moves quickly, listings go stale, and we will get things wrong from time to time. If you spot something inaccurate or out of date, tell us — corrections are logged and reflected in the listing's review date.
Research and drafting on Operator Record are AI-assisted, with human editorial review before publication. That includes this page.
It would be inconsistent to document how other people's AI systems operate while staying vague about our own. So, concretely: listings are compiled by working through the source hierarchy described in our methodology, an evidence level is assigned based on what those sources actually support, and a human is editorially responsible for what appears here. AI assistance doesn't move the evidence standard — nothing gets labeled beyond what its sources justify.
If a listing is wrong, tell us. Corrections are logged and reflected in the last-reviewed date.