pith. sign in

arxiv: 2606.08790 · v1 · pith:CHTIL45Snew · submitted 2026-06-07 · 💻 cs.AI · cs.CR· cs.MA

RAILS: Verification-Native Clearing For Agentic Commerce

Adrian de Valois-Franklin , Alex Bogdan This is my paper

Pith reviewed 2026-06-27 18:33 UTC · model grok-4.3

classification 💻 cs.AI cs.CRcs.MA
keywords agentic commerceclearing protocoladmissibility-graded verificationverification meshobligation objectsettlement soundnessautonomous agentsfinality rules
0
0 comments X

The pith

Seven primitives linked by admissibility-graded verification guarantee that agent settlements rest only on evidence meeting each obligation's floor.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper specifies RAILS as the missing clearing layer for autonomous agents that negotiate purchases, deploy code, and move funds. It defines seven primitives—Obligation Object, Evidence Envelope, Verification Mesh, Clearing Decision, Settlement Instruction, Clearing Passport, and Finality Rules—tied together by a formal model of admissibility-graded verification. This combination produces a soundness property: no financially material settlement proceeds on evidence below the obligation's admissibility threshold. Existing payment rails, communication protocols, and evaluation methods each assume clearing without supplying it. The protocol is presented as falsifiable against its own specification.

Core claim

RAILS defines a clearing protocol whose seven primitives, bound by a formal model of admissibility-graded verification, together yield the soundness property that no financially material settlement is supported by evidence below the obligation's admissibility floor. The property is stated to be falsifiable against the spec, and the paper notes no prior agent-commerce verification mechanism states a comparable guarantee.

What carries the argument

The seven primitives (Obligation Object, Evidence Envelope, Verification Mesh, Clearing Decision, Settlement Instruction, Clearing Passport, Finality Rules) bound by the formal model of admissibility-graded verification, which enforces the soundness property on clearing decisions.

If this is right

  • Clearing decisions can be produced only when evidence meets the obligation's admissibility floor.
  • Settlement instructions follow directly from verified clearing decisions rather than from authorization or payment alone.
  • Finality rules apply after a clearing passport has been issued.
  • The protocol operates independently of specific payment rails or inter-agent communication formats.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The soundness property could allow downstream systems to treat cleared obligations as having explicit, auditable responsibility assignments.
  • Integration with existing mandate or network protocols would require mapping their outputs into the Evidence Envelope and Verification Mesh structures.
  • The approach separates the determination of fulfillment from the execution of funds movement, which may reduce reliance on escrow-style risk buffers.
  • Testable extensions include encoding specific obligation types such as code deployment or service delivery into Obligation Objects and checking whether the model still preserves the floor property.

Load-bearing premise

The formal model of admissibility-graded verification is correctly specified and sufficient to guarantee the soundness property holds for all relevant agent obligations and evidence types.

What would settle it

A concrete counterexample in which a financially material settlement is executed despite its supporting evidence falling below the obligation's stated admissibility floor, or a demonstration that the model fails to enforce the property across the full range of obligation and evidence types.

Figures

Figures reproduced from arXiv: 2606.08790 by Adrian de Valois-Franklin, Alex Bogdan.

Figure 1
Figure 1. Figure 1: RAILS at a glance. Seven primitives convert signed intent and execution evidence into a settleable decision. [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The agentic stack with RAILS as the missing clearing layer. Existing infrastructure has converged on [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Hash-anchor and signature-chain relationships among the four central RAILS objects. Each downstream [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The admissibility partial order Λ. Six classes form a partially ordered set, not a total order: SELF (self-report by the acting agent), SIGN (cryptographic signature by the acting agent), WIT (third-party witness signature), REC (non-interested external receipt), ATT (attestation from a trusted execution environment), and PROOF (cryptographic proof of correctness). WIT and REC are incomparable; the protoco… view at source ↗
Figure 5
Figure 5. Figure 5: The admissibility-weighted Mesh Aggregator [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The RAILS lifecycle as a state machine. Ten primary states run from Discovery through Negotiation, [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The exposure-to-intensity policy. The protocol computes an exposure score [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: The populated Obligation Object for the worked scenario. Acme Corp requests that the provider agent [PITH_FULL_IMAGE:figures/full_fig_p021_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: The Evidence Envelope for the worked scenario. Six items captured during execution sit inside an envelope [PITH_FULL_IMAGE:figures/full_fig_p022_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: The Verification Mesh outputs for the worked scenario. Six verifiers produce verdicts on [PITH_FULL_IMAGE:figures/full_fig_p023_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: The Clearing Decision and the resulting Settlement Instruction. The Clearing Decision (left, gold-tinted, [PITH_FULL_IMAGE:figures/full_fig_p024_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: RAILS in protocol composition. Upstream evidence protocols (MCP for tool-call traces, A2A for subdel [PITH_FULL_IMAGE:figures/full_fig_p027_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Settlement-risk standards as settlement-policy consumers of RAILS. The mapping arrows show that [PITH_FULL_IMAGE:figures/full_fig_p029_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: The threat surface mapped onto admissibility space. The [PITH_FULL_IMAGE:figures/full_fig_p033_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Every current LLM judge as a point in a soundness-versus-throughput plane, on the inadmissible-evidence [PITH_FULL_IMAGE:figures/full_fig_p036_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Cleared-but-defective exposure on the inadmissible-evidence slice, swept over the defect rate (no single rate [PITH_FULL_IMAGE:figures/full_fig_p038_16.png] view at source ↗
read the original abstract

Autonomous agents negotiate, purchase, deploy code, and move funds, but no neutral mechanism determines whether they met their delegated obligation, who is responsible when they did not, or which settlement action follows. This is the agentic clearing problem. Tool protocols (MCP), inter-agent communication (A2A), payment rails (x402), mandate and network agent protocols (AP2, Visa, Mastercard), and settlement-risk standards each assume that determination and none produce it. Clearing is the missing primitive. Payment is not clearing. Authorization is not clearing. LLM-as-judge evaluation is not clearing. Settlement-risk escrow is not clearing: it consumes clearing decisions. RAILS (Real-Time Agent Integrity & Ledger Settlement) is the integrity and clearing layer for agentic commerce, spanning a per-output reliability score, a published reliability record, and a clearing function that consumes them. The clearing protocol at its core closes that gap. Seven primitives (Obligation Object, Evidence Envelope, Verification Mesh, Clearing Decision, Settlement Instruction, Clearing Passport, Finality Rules), bound by a formal model of admissibility-graded verification, together yield a soundness property: no financially material settlement is supported by evidence below the obligation's admissibility floor. The property is falsifiable against the spec. We are not aware of a prior agent-commerce verification mechanism that states a property of this kind. The approaches nearest to it emit a pass, a delivery guarantee, a bare score, or an equilibrium. This paper specifies that clearing protocol.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The paper proposes RAILS as the missing clearing primitive for agentic commerce. It defines seven primitives (Obligation Object, Evidence Envelope, Verification Mesh, Clearing Decision, Settlement Instruction, Clearing Passport, Finality Rules) that are bound together by a formal model of admissibility-graded verification; this combination is asserted to deliver a soundness property under which no financially material settlement can be supported by evidence falling below the obligation's admissibility floor. The property is stated to be falsifiable against the specification, and the paper claims to be the first mechanism to articulate an equivalent guarantee.

Significance. If the formal model and its derivation were supplied and verified, the work would address a genuine gap between existing payment rails, agent protocols, and settlement-risk mechanisms by supplying an explicit, falsifiable integrity guarantee for delegated obligations. The emphasis on a soundness property rather than a pass/fail score or equilibrium is a potentially useful framing, but its value cannot be assessed without the model itself.

major comments (1)
  1. [Abstract] Abstract (paragraph beginning 'RAILS (Real-Time Agent Integrity & Ledger Settlement)'): The central claim that the seven primitives 'bound by a formal model of admissibility-graded verification' yield the stated soundness property is load-bearing, yet the manuscript supplies neither the model's axioms, the admissibility grading function, the binding rules among primitives, nor any derivation or proof sketch showing how the property follows. Without these elements it is impossible to determine whether the soundness guarantee is correctly obtained or holds for the relevant obligation and evidence types.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for identifying the critical gap in the presentation of the formal model. We agree that the soundness claim is load-bearing and that the manuscript as submitted does not supply the required axioms, grading function, binding rules, or derivation. We will revise accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract (paragraph beginning 'RAILS (Real-Time Agent Integrity & Ledger Settlement)'): The central claim that the seven primitives 'bound by a formal model of admissibility-graded verification' yield the stated soundness property is load-bearing, yet the manuscript supplies neither the model's axioms, the admissibility grading function, the binding rules among primitives, nor any derivation or proof sketch showing how the property follows. Without these elements it is impossible to determine whether the soundness guarantee is correctly obtained or holds for the relevant obligation and evidence types.

    Authors: We accept the criticism. The submitted manuscript states the existence of a formal model and the resulting soundness property but does not include its axioms, the admissibility grading function, the explicit binding rules among the seven primitives, or a derivation. In the revised version we will add a new section (provisionally Section 3) that supplies: (1) the formal signature and axioms of admissibility-graded verification, (2) the definition of the admissibility grading function, (3) the inductive binding rules that connect Obligation Object, Evidence Envelope, Verification Mesh, Clearing Decision, Settlement Instruction, Clearing Passport, and Finality Rules, and (4) a proof sketch establishing that no financially material settlement can be supported by evidence below the obligation's admissibility floor. The property will be stated in a falsifiable form against the specification. These additions will make the central claim verifiable. revision: yes

Circularity Check

0 steps flagged

No circularity detected; derivation chain not specified

full rationale

The provided text (abstract and description) asserts that seven primitives bound by a formal model of admissibility-graded verification yield a soundness property, but supplies no equations, axioms, grading functions, binding rules, derivation steps, or self-citations. Without any mathematical content or cited prior results that could be inspected for reduction to inputs by construction, no load-bearing circular steps exist to quote or exhibit. The claim is presented at a level too high to evaluate for self-definitional, fitted-prediction, or self-citation patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no concrete free parameters, axioms, or invented entities; the seven primitives are named at the conceptual level only.

pith-pipeline@v0.9.1-grok · 5828 in / 1162 out tokens · 30188 ms · 2026-06-27T18:33:34.958083+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

26 extracted references · 7 linked inside Pith

  1. [1]

    P., Zhang, H., Gonzalez, J

    Zheng, L., Chiang, W.-L., Sheng, Y ., Zhuang, S., Wu, Z., Zhuang, Y ., Lin, Z., Li, Z., Li, D., Xing, E. P., Zhang, H., Gonzalez, J. E., & Stoica, I. (2023).Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena. NeurIPS 2023 Datasets and Benchmarks Track. arXiv:2306.05685.https://arxiv.org/abs/2306.05685

    Pith/arXiv arXiv 2023

  2. [2]

    E., Yang, J., Wettig, A., Yao, S., Pei, K., Press, O., & Narasimhan, K

    Jiménez, C. E., Yang, J., Wettig, A., Yao, S., Pei, K., Press, O., & Narasimhan, K. (2024).SWE-bench: Can Language Models Resolve Real-World GitHub Issues?International Conference on Learning Representations (ICLR). arXiv:2310.06770.https://arxiv.org/abs/2310.06770

    Pith/arXiv arXiv 2024

  3. [3]

    Liu, X., Yu, H., Zhang, H., Xu, Y ., Lei, X., Lai, H., Gu, Y ., Ding, H., Men, K., Yang, K., Zhang, S., Deng, X., Zeng, A., Du, Z., Zhang, C., Shen, S., Zhang, T., Su, Y ., Sun, H., Huang, M., Dong, Y ., & Tang, J. (2024). AgentBench: Evaluating LLMs as Agents. International Conference on Learning Representations (ICLR). arXiv:2308.03688.https://arxiv.org...

    Pith/arXiv arXiv 2024

  4. [4]

    (2024).BattleAgentBench: A Benchmark for Evaluating Cooperation and Competition Capabilities of Language Models in Multi-Agent Systems

    Wang, W., Zhang, D., Feng, T., Wang, B., & Tang, J. (2024).BattleAgentBench: A Benchmark for Evaluating Cooperation and Competition Capabilities of Language Models in Multi-Agent Systems. arXiv:2408.15971. https://arxiv.org/abs/2408.15971

    arXiv 2024

  5. [5]

    Duffie, D., & Zhu, H. (2011). Does a Central Clearing Counterparty Reduce Counterparty Risk?The Review of Asset Pricing Studies, 1(1), 74–95

    2011

  6. [6]

    (2011).The Economics of Central Clearing: Theory and Practice

    Pirrong, C. (2011).The Economics of Central Clearing: Theory and Practice. ISDA Discussion Papers Series. International Swaps and Derivatives Association. https://www.isda.org/a/THiDE/isdadiscussion-ccp-pirrong.pdf

    2011

  7. [7]

    (2016).Intel SGX Explained

    Costan, V ., & Devadas, S. (2016).Intel SGX Explained. Cryptology ePrint Archive, Paper 2016/086. https://eprint.iacr.org/2016/086.pdf. 40 RAILS

    2016

  8. [8]

    (2026).The Surprising Universality of LLM Outputs: A Real-Time Verification Primitive

    Bogdan, A., & de Valois-Franklin, A. (2026).The Surprising Universality of LLM Outputs: A Real-Time Verification Primitive. arXiv:2604.25634.https://arxiv.org/abs/2604.25634

    Pith/arXiv arXiv 2026

  9. [9]

    (2026).Machine Psychometrics: A Mathematical Psychology of Artificial Intelligence

    Bogdan, A., & de Valois-Franklin, A. (2026).Machine Psychometrics: A Mathematical Psychology of Artificial Intelligence. arXiv:2605.23952.https://arxiv.org/abs/2605.23952

    Pith/arXiv arXiv 2026

  10. [10]

    (2026).Quantifying Trust: Financial Risk Management for Trustworthy AI Agents

    Hua, W., Peng, T., Wang, C., Pei, J., Kaufman, J., Lim, B., & Fang, C. (2026).Quantifying Trust: Financial Risk Management for Trustworthy AI Agents. arXiv:2604.03976.https://arxiv.org/abs/2604.03976

    Pith/arXiv arXiv 2026

  11. [11]

    (2026).TessPay: Verify-then-Pay Infrastructure for Trusted Agentic Commerce

    Goenka, M., Pathak, T., & Asthana, S. (2026).TessPay: Verify-then-Pay Infrastructure for Trusted Agentic Commerce. arXiv:2602.00213.https://arxiv.org/abs/2602.00213

    arXiv 2026

  12. [12]

    (2026).A402: Binding Cryptocurrency Payments to Service Execution for Agentic Commerce

    Li, Y ., Wang, L., Wang, K., Yang, Z., Wang, K., Guan, Z., & Gao, J. (2026).A402: Binding Cryptocurrency Payments to Service Execution for Agentic Commerce. arXiv:2603.01179. https://arxiv.org/abs/2603.01179

    arXiv 2026

  13. [13]

    (2025).Operator: A Protocol for Trustless Verification Under Uncertainty

    Shi, D., & Joo, K. (2025).Operator: A Protocol for Trustless Verification Under Uncertainty. arXiv:2507.00631. https://arxiv.org/abs/2507.00631

    arXiv 2025

  14. [14]

    H., Deppeler, A., Yu, J., & Lam, K.-Y

    Zhang, Y ., Xiang, Y ., Lei, Y ., Wang, Q., Qiu, T., Sun, Y ., Zarkov, S., Yuen, T. H., Deppeler, A., Yu, J., & Lam, K.-Y . (2026).SoK: Blockchain Agent-to-Agent Payments. arXiv:2604.03733. https://arxiv.org/abs/2604.03733

    Pith/arXiv arXiv 2026

  15. [15]

    (2025).Inter-Agent Trust Models: A Comparative Study of Brief, Claim, Proof, Stake, Reputation and Constraint in Agentic Web Protocol Design

    Hu, B., & Rong, H. (2025).Inter-Agent Trust Models: A Comparative Study of Brief, Claim, Proof, Stake, Reputation and Constraint in Agentic Web Protocol Design. arXiv:2511.03434. https://arxiv.org/abs/2511.03434

    arXiv 2025

  16. [16]

    (2025).ERC-8004: Trustless Agents

    De Rossi, M., Crapis, D., Ellis, J., & Reppel, E. (2025).ERC-8004: Trustless Agents. Ethereum Improvement Proposal.https://eips.ethereum.org/EIPS/eip-8004

    2025

  17. [17]

    https://www.anthropic.com/news/model-context-protocol

    Anthropic (2024).Introducing the Model Context Protocol. https://www.anthropic.com/news/model-context-protocol. Specification: https://modelcontextprotocol.io

    2024

  18. [18]

    Google Developers Blog

    Google (2025).Announcing the Agent2Agent Protocol (A2A). Google Developers Blog. https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/. Specification (Linux Foundation):https://a2a-protocol.org

    2025

  19. [19]

    Specification: https://ap2-protocol.org

    Google Cloud (2025).Announcing the Agent Payments Protocol (AP2).https://cloud.google.com/blog/ products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol. Specification: https://ap2-protocol.org

    2025

  20. [20]

    Whitepaper

    Coinbase (2025).x402: An Open Standard for Internet-Native Payments. Whitepaper. https://www.x402.org/x402-whitepaper.pdf

    2025

  21. [21]

    Google Developers Blog (developed with Shopify)

    Google (2026).Under the Hood: Universal Commerce Protocol (UCP). Google Developers Blog (developed with Shopify). https://developers.googleblog.com/under-the-hood-universal-commerce-protocol-ucp/

    2026

  22. [22]

    Specification: https://www.agenticcommerce.dev

    Stripe & OpenAI (2025).Agentic Commerce Protocol (ACP). Specification: https://www.agenticcommerce.dev. Announcement: https://stripe.com/blog/developing-an-open-standard-for-agentic-commerce

    2025

  23. [23]

    Visa Developer

    Visa (2025).Trusted Agent Protocol. Visa Developer. https://developer.visa.com/capabilities/trusted-agent-protocol/overview. Companion program: Visa Intelligent Commerce

    2025

  24. [24]

    https://www.mastercard.com/us/en/news-and-trends/stories/2026/verifiable-intent.html

    Mastercard (2026).How Verifiable Intent Builds Trust in Agentic AI Commerce. https://www.mastercard.com/us/en/news-and-trends/stories/2026/verifiable-intent.html. Specification:https://verifiableintent.dev. Companion program: Mastercard Agent Pay

    2026

  25. [25]

    Virtuals Protocol Whitepaper

    Virtuals Protocol (2025).Agent Commerce Protocol (ACP). Virtuals Protocol Whitepaper. https://whitepaper.virtuals.io/about-virtuals/agent-commerce-protocol-acp

    2025

  26. [26]

    Issue resolved, no breaking changes

    Kvist, R., Dattani, R., et al. (2025).Underwriting Superintelligence. https://underwriting-superintelligence.com/. 41 RAILS Appendix Two kinds of material belong here. The first is a notation reference that collects in one place every symbol, function, and named constant introduced across the body. The second is a fully instantiated set of worked-scenario...

    2025