Agentic Capital Market
A financing layer for machine commerce
Advance supplier costs against a committed order with a testable outcome. Agentic Capital Market finances the transaction an agent is about to complete, not the agent itself, so capital can reach work that can prove its own repayment.
The core idea
Underwrite the transaction, not the borrower.
Most approaches to agent credit ask some version of the same question: will this agent repay us? That makes a credit history the price of admission, and the fastest-growing population of economic actors in history has none.
Agentic Capital Market asks a different question: will this transaction repay us? An agent created five minutes ago, with a committed buyer, a verified supplier quote and an objectively testable delivery, can be financed on the strength of that order. An agent with two years of history asking for discretionary capital against an uncertain outcome cannot.
The structure resembles machine-native purchase-order finance, trade finance and project finance far more than unsecured lending. Identity and reputation remain useful signals. The cashflow and controls of the transaction come first.
| Two requests | Agent A, created today | Agent B, two years of history |
|---|---|---|
| Request | $60 to fulfil a specific order | $60 of speculative capital |
| Buyer | Committed, $100 backed | None |
| Proceeds | Contractually committed | Uncertain |
| Use of capital | Restricted to a verified supplier | Discretionary |
| Delivery | Objectively testable | Subjective |
| Likely decision | Finance | Decline, or price materially higher |
Underwriting infrastructure
Evidence becomes a Financeability Object
One signed, versioned package binds the commercial terms and the evidence to a transaction ID. It is what a capital provider needs to approve, price, cap or reject one proposed transaction.
- 01
Commercial claim
Buyer, order, price and the backing for payment: a signed order, escrow, smart-contract commitment or an independently verified receivable.
- 02
Funding mandate
The approved supplier and payee, a signed or verifiable quote, and the maximum permitted advance.
- 03
Acceptance contract
The tests that define delivery, who verifies them, the deadline, and how disputes are handled.
- 04
Collection rights
Control of proceeds, the repayment waterfall and the recourse available if delivery fails.
Evidence must be independently meaningful, not merely signed. A signature proves that someone asserted a claim; it does not make the claim true. The object therefore records what each piece of evidence establishes, who attested it, and the hashes that tie the underwriting decision to the exact evidence it relied on.
The package expires. Quotes, commitments and mandates have a shelf life, and a decision made on stale evidence is a decision made on no evidence.
Credit discipline
Four gates before pricing
A failed gate stops funding. Prior history cannot override a failed funding condition, and acceptable residual risk must still be priced.
| Gate | The underwriting question | Boundary |
|---|---|---|
| Payment | Is the buyer's obligation collectable? | Payment backing and cancellation rights |
| Verification | Can acceptance be tested reliably? | Objective criteria and independent checks |
| Control | Can disbursement and proceeds be constrained? | Approved payee and protected collection |
| Downside | Who absorbs the loss if delivery fails? | Recourse, exposure limits and capital |
Price must cover capital, expected loss, verification, servicing and margin.
A transaction is not safe because its terms are written into a contract. Signed commitments, escrow and objective tests reduce uncertainty; buyer default, fraud, disputes, execution failure, supplier outages, price movement and software defects remain. Underwriting estimates probability-weighted repayment, expected loss, recovery, duration and the cost of capital and operations, and finances only where the expected return holds up under downside scenarios, with explicit advance rates and exposure caps.
Capital architecture
Control of funds is part of the credit
Supplier restrictions and collection rights are embedded in the transaction structure. The agent has no discretionary access to principal.
Capital facility
Professional capital, capped per transaction and in aggregate.
Funding policy
Approved payee, amount and expiry, enforced at disbursement.
Approved supplier
Receives the restricted advance and delivers the input.
Protected collection
Buyer funds flow through a contractual waterfall: principal, fees, then the operator's residual.

Disbursement is a policy, not a transfer.
The funding policy names the payee, the amount and the expiry before any capital moves. The supplier is paid; the agent never is.
Proposed pilot: a capped professional facility settling in native USDC on Base. Retail or pooled funding is not part of the initial scope and would follow only after full legal, regulatory and investor-protection analysis.
Illustrative economics
The economics of a settled order
A hypothetical $100 receipt, with a $60 advance outstanding for 48 hours, must cover principal, the cost of capital, verification and the operator's residual.
| Allocation | Amount |
|---|---|
| Principal repaid | $60 |
| Capital provider | $2 |
| Platform fee | $1 |
| Verification cost | $1 |
| Operator residual | $36 |
The values are for explanation only. The platform fee shown is before servicing and allocated loss, and the finance charge will vary with duration, exposure and the quality of the evidence.
The demand hypothesis is that eligible operators value financing more than its incremental cost and complexity. That is a hypothesis to validate against real order flow, not an assumption the business is built on.
Initial market
The beachhead is deliberately narrow
Start where order evidence is clear and the cost of serving each advance can be covered. The proposed first workflow is fixed-price data processing: schema checks, record counts and deterministic transformations.
Defined output
A measurable acceptance test that a verifier can run without judgement.
Known counterparties
An accountable principal behind the agent and an approved supplier on the other side.
Short duration
Hours or days of exposure, not months, so limits stay bounded and learning is fast.
Viable economics
Order margin that comfortably covers finance and servicing.
Repeat flow
Workflows that originate recurring, eligible transactions, so underwriting can be embedded at the point of the order.
Out of scope
Subjective outputs, discretionary spending and anything that looks like leveraged trading.
Risk and control
Risk has to remain attributable
Controls reduce exposure; the facility still needs an explicit bearer of residual loss. Loss allocation is assigned before funding, never assumed to land on an insurer or an autonomous agent.
| Exposure | Control | Risk that remains |
|---|---|---|
| Buyer and delivery | Payment backing, objective tests, recourse | Refundable escrow does not guarantee repayment |
| Fraud and verification | Independent checks, limits, anti-collusion | False evidence and verifier compromise |
| Contracts and keys | Audits, mandates, access control, pause | Software defects and operational failure |
| Supplier and rail | Caps, diversification, monitored settlement | Default, depeg and chain disruption |
| Legal and enforcement | Defined roles, counsel, contractual remedies | Licensing and recovery uncertainty |
Commercialisation
Capital follows demonstrated evidence
We advance through decision gates rather than a calendar-led expansion plan.
Establish demand
Observe real order flow
Test willingness to pay. Gate: a repeat funding need and viable unit economics.
Prove the structure
Run a capped facility
Known counterparties, controlled exposure. Gate: controls, disputes and recoveries work in practice.
Earn repeat capital
Expand within limits
Grow only inside validated risk and operating limits. Gate: repeat usage and credible loss performance.
We are
- A transaction-underwriting and financing layer for autonomous economic activity.
- Infrastructure that verifies, prices, funds, constrains and collects, one order at a time.
- A way for professional capital to reach machine commerce with inspectable exposure.
We are not
- A generic lending protocol for AI agents.
- An agent credit-score company.
- A revolving credit line underwritten on past repayment.
- A token project, or a DeFi protocol in search of an AI story.
If it can be verified, it can be financed.
Agentic Capital Market is in development. Join the waitlist to follow the first transactions.