Yesterday the FDIC's proposed reporting forms for stablecoin issuers appeared in the Federal Register. The forms themselves are dull. The reporting cadence is not, and it tells you what supervisors think this instrument actually is.

Weekly is the number that matters

The FDIC published the forms and instructions for supervised permitted payment stablecoin issuers on 20 July, with comments open until 18 September [1]. The collection sits under the proposed rule issued in April and covers both weekly and quarterly returns, with the weekly form built to monitor compliance against the reserve asset requirements [2]. Bank holding companies do not report their balance sheets weekly. Money market funds do. The cadence is the classification.

The tiering is where the operating cost lands

There are two weekly forms. Form PS-01 applies to issuers with $1 billion or more outstanding, or $100 million or more in average daily transaction volume in the prior month. Everyone below both thresholds files the shorter PS-01a [2]. Note the second trigger. An issuer with a modest float but heavy throughput crosses into full reporting on velocity alone, which is precisely the profile of an issuer serving agent traffic. The OCC is running its own forms track for the issuers under its jurisdiction [3].

The wider rulebook is still moving

Market structure remains unsettled. An updated CLARITY Act draft was expected with a possible Senate floor vote in the coming fortnight, though the combined text was not public going into this week and the timeline has slipped repeatedly [4]. Reporting obligations are hardening faster than the statute that surrounds them.

The commercial side did not wait

Cross River extended its Stripe issuing partnership at the start of the month to support agent-initiated card payments, with the agent wallet issuing a restricted single-use virtual card scoped to one transaction so the agent never touches the customer's underlying credentials [5]. On the crypto-native side, roughly 160 million autonomous transactions had cleared the x402 protocol as of early July [6].

Someone has to prove the agent meant to buy it

A-Comm Technologies opened its draft Evidence Protocol for public comment on 13 July, with the window running to 14 August [7]. It aims at one narrow problem: holding a record that establishes an AI-initiated consumer transaction was valid. That is a chargeback question wearing a standards badge. When a customer disputes what their agent bought, someone has to produce evidence of intent, and there is no agreed format yet for what that evidence looks like.

Read from the rails

Weekly reserve reporting is a small phrase with a large operational tail. It means an issuer needs a reserve position it can close, reconcile and attest every seven days, not a month-end process stretched thin. Anyone who has run a daily reconciliation knows the difference between a control that runs monthly and one that runs weekly is not four times the effort. It is a different system, because weekly frequency removes the slack where manual fixes normally hide.

Ten years in payments operations taught me that reporting frequency is how supervisors tell you what they are afraid of. Quarterly says solvency. Weekly says liquidity, and specifically the fear of a redemption run outrunning the last known reserve figure. The velocity threshold makes that explicit: it is not the size of your float that pulls you into full reporting, it is how fast it turns over. If you are building an issuer that settles agent payments, throughput is your product and throughput is also the thing that promotes you into the heavier regime. Worth modelling that before it arrives rather than after.

Reporting cadence is not paperwork. It is the supervisor telling you which failure they expect.

Building in agentic commerce or payments?

This is the intersection I work in. Book thirty minutes and we will scope the build worth doing.

Book an AI consultation