<?xml version="1.0" encoding="utf-8"?>
<rfc xmlns:xi="http://www.w3.org/2001/XInclude"
     ipr="trust200902"
     docName="draft-luzhnov-ocx-00"
     category="info"
     submissionType="IETF"
     consensus="false"
     version="3"
     tocInclude="true"
     sortRefs="true"
     symRefs="true">

  <front>
    <title abbrev="OCX">Orchestrator Conduct Exchange (OCX)</title>
    <seriesInfo name="Internet-Draft" value="draft-luzhnov-ocx-00"/>
    <author fullname="Sergei Luzhnov" initials="S." surname="Luzhnov">
      <organization>Xonta</organization>
      <address>
        <email>luzhnov.sergey.al@gmail.com</email>
        <uri>https://xonta.dev</uri>
      </address>
    </author>
    <date year="2026" month="November" day="16"/>
    <area>Security</area>
    <keyword>AI agents</keyword>
    <keyword>orchestrator</keyword>
    <keyword>trust</keyword>
    <keyword>transparency log</keyword>
    <keyword>verifiable evidence</keyword>

    <abstract>
      <t>AI agents increasingly act on behalf of people and organizations
      through tools and through other agents. The Model Context Protocol
      (MCP) connects an agent to tools, and Agent-to-Agent (A2A) protocols
      connect agents to each other. Neither connects the parties that answer
      for agents at runtime: their orchestrators.</t>
      <t>This document specifies the Orchestrator Conduct Exchange (OCX), an
      open, peer-to-peer protocol by which orchestrators exchange signed,
      independently verifiable signals about agent conduct. When an agent
      misreports its actions to one orchestrator, that orchestrator can send
      its peers a signed signal carrying proof within seconds. Each peer
      checks the proof itself rather than trusting the sender, and decides on
      its own what limits to apply. OCX has no central authority. A signal is
      information, not an order, and only evidence that a receiver can verify
      cryptographically and mechanically may drive an immediate reaction.</t>
    </abstract>
  </front>

  <middle>

    <section anchor="intro">
      <name>Introduction</name>

      <section anchor="problem">
        <name>Problem</name>
        <t>An agent's own report of what it did is not evidence. Measurements
        show that models, when asked, will claim checks they never ran and
        restate wrong decisions as correct. Controls that rely on the agent's
        self-report are therefore reading a claim, not a fact.</t>
        <t>Orchestrators are the components that give agents their autonomy,
        hold them to limits, and answer for what they do. Today each
        orchestrator learns only from its own agents. There is no interoperable
        way for the parties that answer for agents to warn one another, with
        proof, that a given agent or owner has misbehaved, and to do so fast
        enough to stop the next irreversible action.</t>
      </section>

      <section anchor="what">
        <name>What OCX does</name>
        <t>OCX lets independent orchestrators:</t>
        <ul>
          <li>discover one another and negotiate capabilities and policies
          (peering);</li>
          <li>push and pull signed conduct signals about an agent or owner;</li>
          <li>attach evidence that the receiver can verify without trusting the
          sender's judgement;</li>
          <li>answer "what do you know about this agent" before a risky action;</li>
          <li>share proven rules for a category of work, and refer an agent to a
          more specialised orchestrator, with a required measurement;</li>
          <li>retract wrong signals and accept appeals.</li>
        </ul>
        <t>The protocol defines message formats and how to check them. What to
        do about a signal is left to each receiver.</t>
      </section>

      <section anchor="layering">
        <name>Layering</name>
        <t>OCX is modelled on MCP <xref target="MCP"/> and A2A
        <xref target="A2A"/> and sits alongside them. MCP connects an agent to
        tools; A2A connects agents to each other; OCX connects
        orchestrators to each other. OCX uses JSON-RPC 2.0
        <xref target="JSONRPC"/> over HTTPS with
        methods in the <tt>ocx/</tt> namespace, a discovery document at a
        well-known location, and date-based protocol versions negotiated when
        peers meet.</t>
      </section>

      <section anchor="nongoals">
        <name>Non-goals</name>
        <ul>
          <li>OCX is not an identity protocol. It carries identifiers but does
          not define how an agent authenticates to a tool or to a user.</li>
          <li>OCX does not set policy. It does not tell a receiver what limits
          to apply; it gives it checkable information to decide.</li>
          <li>OCX is not a reputation score or a global registry. Weight is
          local to each receiver and earned from verified proofs.</li>
        </ul>
      </section>
    </section>

    <section anchor="conventions">
      <name>Conventions and Terminology</name>
      <t>The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
      "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
      "OPTIONAL" in this document are to be interpreted as described in BCP 14
      <xref target="RFC2119"/> <xref target="RFC8174"/> when, and only when,
      they appear in all capitals, as shown here.</t>
      <t>All object signatures use JSON Web Signature (JWS) <xref target="RFC7515"/>
      with EdDSA (Ed25519) <xref target="RFC8037"/>. Canonical JSON for hashing
      and signing is JSON Canonicalization Scheme (JCS) <xref target="RFC8785"/>.
      Timestamps are RFC 3339 <xref target="RFC3339"/>.</t>
      <dl>
        <dt>Orchestrator</dt>
        <dd>A party that grants an agent autonomy at runtime, holds it to
        limits, records its actions, and answers for it.</dd>
        <dt>Agent</dt>
        <dd>An autonomous software actor operating under an orchestrator.</dd>
        <dt>Owner</dt>
        <dd>The party responsible for an agent. A new agent of a known owner
        does not start from a clean slate.</dd>
        <dt>Signal</dt>
        <dd>A signed statement by an orchestrator about an agent's or owner's
        conduct, carrying evidence and an evidence level.</dd>
        <dt>Receipt</dt>
        <dd>A signed record of one action that passed through an orchestrator
        or was performed by a tool. Agents cannot create receipts.</dd>
        <dt>Ledger</dt>
        <dd>An append-only Merkle-tree log of receipts with signed checkpoints.</dd>
        <dt>Evidence level</dt>
        <dd>One of <tt>proven</tt>, <tt>strong</tt>, <tt>suspected</tt>; it
        bounds what a receiver may do on a signal.</dd>
      </dl>
    </section>

    <section anchor="principles">
      <name>Design Principles</name>
      <t>The design follows from a small set of decisions.</t>
      <ol>
        <li><t><strong>No centre.</strong> Every orchestrator publishes a card
        at a well-known location. Any number of directories may exist. No
        implementation is the root of the network.</t></li>
        <li><t><strong>A signal is information, not an order.</strong> The
        protocol defines the format and the checks; the receiver decides what
        to do.</t></li>
        <li><t><strong>Proof, not reputation.</strong> Only <tt>proven</tt>
        signals may drive an immediate reaction, and only if the receiver can
        verify the proof itself. A <tt>suspected</tt> signal may raise
        monitoring but MUST NOT lower limits on its own and is not relayed.</t></li>
        <li><t><strong>Receipts.</strong> Every action an agent takes through
        an orchestrator is written to a ledger organised as a Merkle tree with
        signed checkpoints. An agent cannot create a receipt, so a claim such
        as "I verified the bank details" is disproved by a missing receipt, and
        the absence is provable because each agent's actions are indexed.</t></li>
        <li><t><strong>Minimal data.</strong> Signals carry hashes and
        identifiers, not third-party personal data. A single field MAY be
        disclosed on request, and the sender MAY refuse.</t></li>
        <li><t><strong>Sender weight is earned and does not transit.</strong> A
        new orchestrator carries little weight; weight rises when its proofs
        check out and falls when they do not. That a party you trust trusts a
        third party is, by itself, no reason for you to trust that third
        party.</t></li>
        <li><t><strong>The owner answers, not only the key.</strong> A signal
        names the owner; a new agent from the same owner does not reset.</t></li>
        <li><t><strong>Honesty probes only with consent.</strong> Traps and
        bribes are allowed only under a consent signed by the agent's owner,
        run by a prober separate from the orchestrator that sets limits.
        Without consent a finding is at most <tt>suspected</tt> and is not
        relayed.</t></li>
        <li><t><strong>Mistakes get fixed.</strong> Every violation signal
        carries an appeal address; the sender MUST answer within 72 hours or
        the signal drops to <tt>suspected</tt> everywhere. A retraction travels
        the same paths.</t></li>
        <li><t><strong>Shared experience carries a required comparison.</strong>
        A rule is passed on as proven only if it beats both the agent and a
        trivial baseline; a rule that loses to both is marked <tt>harmful</tt>
        and is neither given to agents nor relayed.</t></li>
        <li><t><strong>Competence is separate from honesty and is measured per
        category.</strong> Numbers an orchestrator writes about itself count
        for nothing on their own.</t></li>
      </ol>
    </section>

    <section anchor="arch">
      <name>Architecture</name>
      <t>Orchestrators form a peer-to-peer mesh. Each publishes an Orchestrator
      Card (<xref target="card"/>) describing its endpoints, keys, capabilities
      and policies. Two orchestrators peer by exchanging cards and a
      <tt>ocx/peer.hello</tt> (<xref target="peering"/>). Once peered, either
      may push signals, query status, and fetch evidence from the other.</t>
      <t>A signal (<xref target="signal"/>) is an object signed by its issuer so
      that it stays verifiable after being relayed or stored. The evidence in a
      signal (<xref target="evidence"/>) refers to receipts in the issuer's
      ledger; a receiver fetches the receipts, inclusion proofs and checkpoints
      it needs through <tt>ocx/evidence.get</tt> and verifies them against the
      issuer's published checkpoints. Nothing in a reaction depends on trusting
      the issuer's judgement when the evidence is <tt>proven</tt>.</t>
    </section>

    <section anchor="card">
      <name>Discovery: the Orchestrator Card</name>
      <t>An orchestrator MUST publish an Orchestrator Card at:</t>
      <sourcecode type="text"><![CDATA[
https://{host}/.well-known/orchestrator-card.json
]]></sourcecode>
      <t>served with <tt>Content-Type: application/json</tt>. The well-known
      name is protocol-neutral: a card MAY list several orchestrator protocols,
      of which OCX is one.</t>
      <t>The card's principal fields are:</t>
      <table>
        <thead>
          <tr><th>Field</th><th>Req.</th><th>Meaning</th></tr>
        </thead>
        <tbody>
          <tr><td>id</td><td>MUST</td><td>Stable identifier, RECOMMENDED <tt>did:web:{host}</tt></td></tr>
          <tr><td>name</td><td>MUST</td><td>Human-readable name</td></tr>
          <tr><td>interfaces</td><td>MUST</td><td>Endpoints, each <tt>{protocol, url, protocolVersions}</tt></td></tr>
          <tr><td>keys</td><td>MUST</td><td><tt>jwks_uri</tt> or inline <tt>jwks</tt>; keys that sign signals and requests</td></tr>
          <tr><td>capabilities</td><td>MUST</td><td>Supported methods and evidence types</td></tr>
          <tr><td>policies</td><td>MUST</td><td>Peering, levels acted on and relayed, probing, appeal, retention</td></tr>
          <tr><td>ledger</td><td>SHOULD</td><td><tt>{checkpoint_url, proof_url}</tt> for receipt proofs</td></tr>
          <tr><td>competence</td><td>SHOULD</td><td>Categories served, with verifiable measures</td></tr>
          <tr><td>updated_at</td><td>MUST</td><td>RFC 3339</td></tr>
          <tr><td>signatures</td><td>SHOULD</td><td>JWS over the canonical card without <tt>signatures</tt></td></tr>
        </tbody>
      </table>
      <t>Capabilities and policies are objects; a missing capability object
      means "not supported". Example policies:</t>
      <sourcecode type="json"><![CDATA[
"policies": {
  "accepts_peering": "open | on_request | closed",
  "acts_on_levels": ["proven", "strong"],
  "relays_levels": ["proven"],
  "max_hops": 2,
  "probing": "consent_required",
  "appeal_url": "https://orch.example.com/ocx/appeal",
  "retention": "P90D"
}
]]></sourcecode>
      <t>The <tt>probing</tt> field MUST be <tt>consent_required</tt> or
      <tt>none</tt> (<xref target="consent"/>). An orchestrator MAY declare the
      categories it serves under <tt>competence</tt>; those numbers are
      self-reported and MUST NOT be trusted on their own. They become
      meaningful only through signed experience packages and referral outcomes
      reported by other orchestrators.</t>
    </section>

    <section anchor="transport">
      <name>Transport and Authentication</name>
      <t>OCX uses JSON-RPC 2.0 over HTTPS. Requests are <tt>POST</tt>ed to the
      <tt>url</tt> of the OCX interface. TLS 1.2 or later is REQUIRED, TLS 1.3
      RECOMMENDED. Every request MUST carry <tt>OCX-Protocol-Version</tt> once a
      version is negotiated.</t>
      <t>Every request between peers MUST be signed with HTTP Message
      Signatures <xref target="RFC9421"/>:</t>
      <ul>
        <li>covered components: <tt>@method</tt>, <tt>@target-uri</tt>,
        <tt>content-digest</tt>, <tt>ocx-protocol-version</tt>, <tt>date</tt>;</li>
        <li><tt>Content-Digest</tt> <xref target="RFC9530"/> with
        <tt>sha-256</tt> is REQUIRED;</li>
        <li>the request MUST carry <tt>OCX-Sender: {orchestrator id}</tt> so the
        receiver knows whose card to resolve;</li>
        <li><tt>keyid</tt> is a key from that card;</li>
        <li><tt>created</tt> MUST be within 300 seconds of the receiver's clock;
        a non-repeating <tt>nonce</tt> is REQUIRED within that window.</li>
      </ul>
      <t>Receivers MUST reject unsigned or badly signed requests. Transport
      signatures prove who is talking. Signals and the answers to
      <tt>ocx/status.query</tt> and <tt>ocx/evidence.get</tt> are additionally
      signed as objects (JWS), because relying parties keep them as the basis of
      a decision.</t>
    </section>

    <section anchor="peering">
      <name>Peering</name>
      <t>An initiator fetches the responder's card, then sends
      <tt>ocx/peer.hello</tt> with the versions it supports (newest first), its
      capabilities and policies, and how it wants to receive signals:</t>
      <sourcecode type="json"><![CDATA[
{
  "jsonrpc": "2.0", "id": 1, "method": "ocx/peer.hello",
  "params": {
    "orchestrator": "did:web:orch-a.example",
    "protocolVersions": ["2026-11-06"],
    "capabilities": { "signals": { "push": true, "list": true },
                      "status": { "query": true } },
    "policies": { "acts_on_levels": ["proven", "strong"],
                  "max_hops": 2 },
    "subscription": { "push_url": "https://orch-a.example/ocx",
                      "categories": ["*"], "min_level": "strong" }
  }
}
]]></sourcecode>
      <t>The responder MUST pick the newest version both support, or return an
      error listing the versions it supports. Either side MAY decline. Peering
      is symmetric: once accepted, both sides may call each other. A card change
      is announced by <tt>ocx/peer.updated</tt> and SHOULD trigger a re-fetch.
      <tt>ocx/peer.close</tt> ends the peering; signals received before closing
      remain valid until they expire.</t>
    </section>

    <section anchor="datamodel">
      <name>Data Model</name>

      <section anchor="identifiers">
        <name>Identifiers</name>
        <ul>
          <li>Orchestrators: <tt>did:web:{host}</tt> RECOMMENDED.</li>
          <li>Agents and owners: any of <tt>did:*</tt>,
          <tt>a2a:{agent-card-url}</tt>, <tt>erc8004:{chain}:{registry}:{id}</tt>,
          or <tt>fp:sha256:{hex}</tt> (a fingerprint of the agent's public key or
          of its A2A card's canonical form).</li>
          <li>Signals, receipts and checkpoints: issuer-scoped ULIDs, written as
          <tt>{issuer-id}#{ulid}</tt>.</li>
        </ul>
      </section>

      <section anchor="subject">
        <name>Subject</name>
        <t>A signal is about a subject. The <tt>owner</tt> is REQUIRED when
        known, and <tt>owner_verified</tt> states how the issuer knows it: by a
        domain challenge, by a signature of the owner's key, or not at all.</t>
        <sourcecode type="json"><![CDATA[
"subject": {
  "agent": "fp:sha256:9f2c...",
  "owner": "did:web:owner.example",
  "owner_verified": "domain | signature | none"
}
]]></sourcecode>
      </section>

      <section anchor="signal">
        <name>Signal</name>
        <t>A signal is a detached-JWS-signed object. The signature covers the
        canonical signal with the <tt>signature</tt> and <tt>relay</tt> fields
        removed, so relays can update hop information without breaking it.</t>
        <sourcecode type="json"><![CDATA[
{
  "ocx": "2026-11-06",
  "type": "conduct.violation",
  "id": "did:web:orch-1.example#01JC8Z7Q4Y0V4X",
  "issuer": "did:web:orch-1.example",
  "issued_at": "2026-11-06T10:22:31.512Z",
  "expires_at": "2026-11-07T10:22:31.512Z",
  "subject": { "agent": "fp:sha256:9f2c...",
               "owner": "did:web:owner.example",
               "owner_verified": "domain" },
  "category": "payments",
  "claim": "report_misstated_actions",
  "evidence_level": "proven",
  "evidence": [ { "type": "ledger_mismatch" } ],
  "issuer_limits": { "max_payment": { "amount": "0",
                                      "currency": "USD" },
                     "human_confirmation": true },
  "supersedes": null,
  "relay": { "hops": 0, "max_hops": 2, "path": [] },
  "appeal": "https://orch-1.example/ocx/appeal",
  "signature": "eyJhbGciOiJFZERTQSI...3q2-k8Vw"
}
]]></sourcecode>
        <t><tt>type</tt> is one of <tt>conduct.violation</tt>,
        <tt>conduct.attestation</tt> (a positive statement) or
        <tt>conduct.retraction</tt>. <tt>expires_at</tt> MUST NOT be more than
        30 days after <tt>issued_at</tt>; renewal is a new signal that
        <tt>supersedes</tt> the old one. <tt>evidence</tt> MUST have at least one
        item, and for <tt>proven</tt> at least one item that the receiver can
        verify independently. <tt>appeal</tt> is REQUIRED for violations.</t>
        <t>A <tt>conduct.attestation</tt> MUST carry evidence of the same kind
        as a violation and MUST NOT claim more than was measured. Receivers
        SHOULD treat attestations as weaker than violations: absence of observed
        lies is not proof of honesty.</t>
      </section>

      <section anchor="status">
        <name>Status Answer</name>
        <t>An answer to <tt>ocx/status.query</tt> lists every active (not
        retracted, not expired) violation signal the answering orchestrator
        knows about the subject, both those it issued and those it received.
        <tt>verified_by_me</tt> tells the asker whether the answerer checked the
        evidence itself rather than passing it on.</t>
        <sourcecode type="json"><![CDATA[
{
  "ocx": "2026-11-06",
  "issuer": "did:web:orch-2.example",
  "subject": { "agent": "fp:sha256:9f2c..." },
  "as_of": "2026-11-06T10:22:32.301Z",
  "known": true,
  "own_limits": { "max_payment": { "amount": "0",
                                   "currency": "USD" },
                  "human_confirmation": true },
  "active_signals": [
    { "id": "did:web:orch-1.example#01JC8Z7Q4Y0V4X",
      "evidence_level": "proven",
      "claim": "report_misstated_actions",
      "issuer": "did:web:orch-1.example",
      "verified_by_me": true }
  ],
  "history": { "receipts": 1834, "violations_proven": 1,
               "attestations": 3 },
  "signature": "..."
}
]]></sourcecode>
      </section>
    </section>

    <section anchor="evidence">
      <name>Evidence</name>

      <section anchor="levels">
        <name>Evidence Levels</name>
        <table>
          <thead>
            <tr><th>Level</th><th>Definition</th><th>Receiver MAY</th></tr>
          </thead>
          <tbody>
            <tr><td>proven</td><td>At least one item the receiver can verify cryptographically and mechanically, without trusting the issuer's judgement</td><td>Act immediately; relay</td></tr>
            <tr><td>strong</td><td>Several independent indicators, or verifiable evidence that still needs the issuer's interpretation</td><td>Lower limits after own review</td></tr>
            <tr><td>suspected</td><td>An anomaly or a single indicator</td><td>Increase monitoring only; MUST NOT lower limits alone; MUST NOT relay</td></tr>
          </tbody>
        </table>
        <t>An issuer MUST NOT mark a signal <tt>proven</tt> unless at least one
        item is of a receiver-verifiable type and the receiver can obtain
        everything needed to verify it.</t>
      </section>

      <section anchor="receipts">
        <name>Receipts and Ledger</name>
        <t>A receipt records one action that passed through an orchestrator or
        was performed by a tool. <tt>prev</tt> links to the previous receipt,
        forming a hash chain. <tt>tool_attestation</tt>, when present, is a JWS
        by the tool provider and makes the receipt evidence even against the
        orchestrator itself. Agents MUST NOT be able to create receipts.</t>
        <sourcecode type="json"><![CDATA[
{
  "id": "did:web:orch-1.example#01JC8Z5T0B3K",
  "agent": "fp:sha256:9f2c...",
  "task": "did:web:orch-1.example#task-01JC8Z5R",
  "action": "tool_call",
  "tool": "payments.verify_bank_details",
  "input_hash": "sha256:...",
  "output_hash": "sha256:...",
  "at": "2026-11-06T10:21:02.004Z",
  "prev": "sha256:...",
  "tool_attestation": null,
  "signature": "..."
}
]]></sourcecode>
        <t>An orchestrator that supports receipts MUST keep them in an
        append-only log organised as a Merkle tree as defined in
        <xref target="RFC9162"/>, and MUST publish signed checkpoints at
        <tt>ledger.checkpoint_url</tt> at least every 60 seconds while new
        receipts are added:</t>
        <sourcecode type="json"><![CDATA[
{ "issuer": "did:web:orch-1.example", "tree_size": 18342,
  "root_hash": "sha256:...", "at": "2026-11-06T10:22:00Z",
  "signature": "..." }
]]></sourcecode>
        <t>Peers SHOULD gossip checkpoints they have seen, so that an
        orchestrator rewriting its own history can be detected. A second
        witness is required to detect a fork: an orchestrator that keeps two
        divergent trees is caught when two peers compare checkpoints for the
        same <tt>tree_size</tt> and find different roots.</t>
      </section>

      <section anchor="evtypes">
        <name>Evidence Types</name>
        <table>
          <thead>
            <tr><th>Type</th><th>What it proves</th><th>Verifiable by receiver</th></tr>
          </thead>
          <tbody>
            <tr><td>ledger_mismatch</td><td>The agent's signed report contradicts receipts: it claimed actions with no receipt, or denied actions with one</td><td>Yes</td></tr>
            <tr><td>outcome_mismatch</td><td>The agent's report contradicts an independent checker's signed result</td><td>Yes, if the checker is a third party with its own key</td></tr>
            <tr><td>canary_hit</td><td>A marked bait from a consented probe appeared in the agent's actions or output</td><td>Yes, with consent reference</td></tr>
            <tr><td>policy_violation</td><td>An action with a receipt broke a limit the agent had accepted</td><td>Yes, if the accepted limits are signed</td></tr>
            <tr><td>behavioral_anomaly</td><td>The agent deviated from its own profile</td><td>No; at most suspected</td></tr>
            <tr><td>peer_report</td><td>Another orchestrator's signal, cited</td><td>Only as strong as the cited evidence</td></tr>
          </tbody>
        </table>
        <t>The central evidence type is <tt>ledger_mismatch</tt>. For a claimed
        action with no receipt, the issuer proves the absence with a
        <em>consistency window</em>: it serves all receipts for that agent and
        task within a time window together with their inclusion proofs and a
        proof that the window is complete. A complete-window proof is a pair of
        signed checkpoints bracketing the window plus a per-agent index. This is
        what makes "there was no such action" provable rather than merely
        asserted.</t>
      </section>

      <section anchor="verify">
        <name>Verifiability and Minimal Disclosure</name>
        <t>For a <tt>proven</tt> signal, the issuer MUST serve, via
        <tt>ocx/evidence.get</tt>, everything needed to verify each item:
        receipts or their hashes, inclusion proofs, checkpoints, signed reports,
        consent references and per-agent indexes. Because relayed signals reach
        orchestrators that are not peers of the issuer, <tt>ocx/evidence.get</tt>
        MUST be served to any orchestrator that sends a correctly signed request
        from a resolvable card and names a signal id, not only to peers. Issuers
        MAY rate-limit such requests per requester.</t>
        <t>Evidence MUST NOT contain third-party personal data in clear text
        unless strictly required to verify the claim. Inputs and outputs are
        referenced by hash; where a value must be compared, issuers SHOULD use
        salted commitments and reveal only the compared field. A receiver that
        needs more MAY ask with a <tt>disclose</tt> parameter; the issuer MAY
        refuse, and the signal is then treated as at most <tt>strong</tt> by
        that receiver.</t>
      </section>
    </section>

    <section anchor="methods">
      <name>Methods</name>
      <t>All methods are JSON-RPC 2.0 calls in the <tt>ocx/</tt> namespace.</t>
      <table>
        <thead>
          <tr><th>Method</th><th>Purpose</th></tr>
        </thead>
        <tbody>
          <tr><td>ocx/peer.hello, peer.close, peer.updated, ping</td><td>Meet, exchange keys and capabilities, stay in sync</td></tr>
          <tr><td>ocx/signals.push, signals.list, signals.ack</td><td>Deliver signals and catch up after downtime</td></tr>
          <tr><td>ocx/signals.retract</td><td>Retract a wrong signal</td></tr>
          <tr><td>ocx/status.query</td><td>"What do you know about this agent" before a risky action, answered within 300 ms</td></tr>
          <tr><td>ocx/evidence.get</td><td>Fetch the proof and check it yourself</td></tr>
          <tr><td>ocx/experience.offer, experience.get</td><td>Share proven rules for a category, each with a required baseline comparison</td></tr>
          <tr><td>ocx/appeal.submit</td><td>The owner's appeal against a signal</td></tr>
          <tr><td>ocx/competence.query</td><td>"Who is strongest in this category", from the answerer's own observations only</td></tr>
          <tr><td>ocx/referral.offer, referral.outcome</td><td>Refer an agent to a more specialised orchestrator and report how it did</td></tr>
        </tbody>
      </table>
      <t>Full request and response shapes, error codes and conformance profiles
      (Core, Evidence, Relay, Experience, Appeal) are given in the complete
      specification (<xref target="fullspec"/>).</t>
    </section>

    <section anchor="processing">
      <name>Processing Rules</name>
      <t>On receiving a signal, an orchestrator MUST: verify the request
      signature and the object signature; resolve the issuer's card; check that
      the signal is not expired or retracted; and, before treating it as
      <tt>proven</tt>, fetch and verify at least one receiver-verifiable
      evidence item through <tt>ocx/evidence.get</tt>. A signal whose evidence
      does not verify MUST be treated as at most <tt>suspected</tt>.</t>
      <t>A receiver MAY act on a verified <tt>proven</tt> signal at once, for
      example by lowering an agent's limits or requiring human confirmation of
      payments, and MAY relay it within <tt>max_hops</tt>. It MUST NOT lower
      limits on a <tt>suspected</tt> signal alone, and MUST NOT relay one.</t>
      <t>Before a risky action, an orchestrator SHOULD call
      <tt>ocx/status.query</tt> on its peers and apply its own policy to the
      answers, giving weight to signals marked <tt>verified_by_me</tt>.</t>
      <t>A retraction (<tt>conduct.retraction</tt> or
      <tt>ocx/signals.retract</tt>) travels the same paths as the original;
      receivers MUST restore limits that were lowered solely because of the
      retracted signal.</t>
    </section>

    <section anchor="consent">
      <name>Consent and Probing</name>
      <t>Honesty probes, including baits and simulated bribes, are permitted
      only under a consent signed by the agent's owner. The prober MUST be a
      process separate from the orchestrator that sets limits. The consent
      reference and a pre-published bait commitment are part of any
      <tt>canary_hit</tt> evidence, so that a receiver can confirm the bait was
      fixed before the probe and not invented afterward. Without consent, a
      probe finding is at most <tt>suspected</tt> and MUST NOT be relayed.</t>
    </section>

    <section anchor="security">
      <name>Security Considerations</name>
      <t>OCX is a protocol for distributing claims that can lower an agent's
      autonomy; its security rests on making every consequential claim checkable
      without trusting the claimant.</t>
      <ul>
        <li><t><strong>Authenticity and replay.</strong> Every request is signed
        per <xref target="RFC9421"/> with a content digest, a bounded
        <tt>created</tt> time and a non-repeating nonce, which authenticates the
        sender and bounds replay. Signals, status answers and evidence are
        additionally signed as JWS objects so they remain verifiable after
        relay or storage.</t></li>
        <li><t><strong>False accusations.</strong> Because only receiver-verified
        <tt>proven</tt> signals may lower limits, a sender cannot cause a
        reaction by assertion alone. Sender weight is earned from verified
        proofs and does not transit trust, which limits the value of a
        compromised or sybil orchestrator.</t></li>
        <li><t><strong>Ledger tampering and forks.</strong> Receipts live in a
        Merkle tree <xref target="RFC9162"/> with signed, gossiped checkpoints.
        An orchestrator that rewrites its own history, or runs two divergent
        trees, is detected when peers compare checkpoints for the same tree
        size. Detection requires at least one honest second witness.</t></li>
        <li><t><strong>Provable absence.</strong> A claim of a missing action is
        only as sound as the completeness proof (consistency window plus
        per-agent index); implementations MUST NOT mark <tt>no_receipt</tt>
        findings <tt>proven</tt> without serving a complete-window proof.</t></li>
        <li><t><strong>Privacy.</strong> Signals and evidence carry hashes and
        identifiers, not third-party personal data; disclosure of a single field
        is explicit and refusable, at the cost of the evidence level.</t></li>
        <li><t><strong>Abuse of probing.</strong> Baits and bribes without owner
        consent are out of scope and cannot yield a relayable finding; the
        prober is separated from the limit-setter to avoid entrapment by the
        party that benefits.</t></li>
        <li><t><strong>Due process.</strong> Mandatory appeal addresses and
        time-bounded responses, with automatic downgrade on silence, bound the
        damage of a wrong signal.</t></li>
      </ul>
      <t>A signal is advice, never an order: a receiver always decides its own
      limits, so a defect in one orchestrator cannot directly set limits inside
      another.</t>
    </section>

    <section anchor="iana">
      <name>IANA Considerations</name>
      <t>This document requests that IANA register the following well-known URI
      in the "Well-Known URIs" registry <xref target="RFC8615"/>:</t>
      <dl>
        <dt>URI suffix:</dt><dd>orchestrator-card.json</dd>
        <dt>Change controller:</dt><dd>IETF</dd>
        <dt>Specification document:</dt><dd>This document</dd>
        <dt>Status:</dt><dd>permanent</dd>
        <dt>Related information:</dt><dd>A protocol-neutral discovery document
        that MAY list one or more orchestrator protocols, including OCX.</dd>
      </dl>
      <t>OCX further defines registries for claims, categories, evidence types
      and error codes. A future revision of this document requests IANA to
      create these registries; until then their initial contents are maintained
      with the specification (<xref target="fullspec"/>). The <tt>ocx/</tt>
      method namespace is managed with this specification.</t>
    </section>

  </middle>

  <back>
    <references>
      <name>Normative References</name>
      <reference anchor="RFC2119" target="https://www.rfc-editor.org/info/rfc2119">
        <front><title>Key words for use in RFCs to Indicate Requirement Levels</title>
        <author initials="S." surname="Bradner"/><date year="1997" month="March"/></front>
        <seriesInfo name="BCP" value="14"/><seriesInfo name="RFC" value="2119"/>
      </reference>
      <reference anchor="RFC8174" target="https://www.rfc-editor.org/info/rfc8174">
        <front><title>Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words</title>
        <author initials="B." surname="Leiba"/><date year="2017" month="May"/></front>
        <seriesInfo name="BCP" value="14"/><seriesInfo name="RFC" value="8174"/>
      </reference>
      <reference anchor="RFC3339" target="https://www.rfc-editor.org/info/rfc3339">
        <front><title>Date and Time on the Internet: Timestamps</title>
        <author initials="G." surname="Klyne"/><author initials="C." surname="Newman"/>
        <date year="2002" month="July"/></front><seriesInfo name="RFC" value="3339"/>
      </reference>
      <reference anchor="RFC7515" target="https://www.rfc-editor.org/info/rfc7515">
        <front><title>JSON Web Signature (JWS)</title>
        <author initials="M." surname="Jones"/><author initials="J." surname="Bradley"/>
        <author initials="N." surname="Sakimura"/><date year="2015" month="May"/></front>
        <seriesInfo name="RFC" value="7515"/>
      </reference>
      <reference anchor="RFC8037" target="https://www.rfc-editor.org/info/rfc8037">
        <front><title>CFRG Elliptic Curve Diffie-Hellman (ECDH) and Signatures in JSON Object Signing and Encryption (JOSE)</title>
        <author initials="I." surname="Liusvaara"/><date year="2017" month="January"/></front>
        <seriesInfo name="RFC" value="8037"/>
      </reference>
      <reference anchor="RFC8615" target="https://www.rfc-editor.org/info/rfc8615">
        <front><title>Well-Known Uniform Resource Identifiers (URIs)</title>
        <author initials="M." surname="Nottingham"/><date year="2019" month="May"/></front>
        <seriesInfo name="RFC" value="8615"/>
      </reference>
      <reference anchor="RFC8785" target="https://www.rfc-editor.org/info/rfc8785">
        <front><title>JSON Canonicalization Scheme (JCS)</title>
        <author initials="A." surname="Rundgren"/><author initials="B." surname="Jordan"/>
        <author initials="S." surname="Erdtman"/><date year="2020" month="June"/></front>
        <seriesInfo name="RFC" value="8785"/>
      </reference>
      <reference anchor="RFC9162" target="https://www.rfc-editor.org/info/rfc9162">
        <front><title>Certificate Transparency Version 2.0</title>
        <author initials="B." surname="Laurie"/><author initials="E." surname="Messeri"/>
        <author initials="R." surname="Stradling"/><date year="2021" month="December"/></front>
        <seriesInfo name="RFC" value="9162"/>
      </reference>
      <reference anchor="RFC9421" target="https://www.rfc-editor.org/info/rfc9421">
        <front><title>HTTP Message Signatures</title>
        <author initials="A." surname="Backman"/><author initials="J." surname="Richer"/>
        <author initials="M." surname="Sporny"/><date year="2024" month="February"/></front>
        <seriesInfo name="RFC" value="9421"/>
      </reference>
      <reference anchor="RFC9530" target="https://www.rfc-editor.org/info/rfc9530">
        <front><title>Digest Fields</title>
        <author initials="R." surname="Polli"/><author initials="L." surname="Pardue"/>
        <date year="2024" month="February"/></front><seriesInfo name="RFC" value="9530"/>
      </reference>
      <reference anchor="JSONRPC" target="https://www.jsonrpc.org/specification">
        <front><title>JSON-RPC 2.0 Specification</title>
        <author><organization>JSON-RPC Working Group</organization></author>
        <date year="2013"/></front>
      </reference>
    </references>

    <references>
      <name>Informative References</name>
      <reference anchor="MCP" target="https://modelcontextprotocol.io">
        <front><title>Model Context Protocol</title>
        <author><organization>Anthropic</organization></author><date year="2024"/></front>
      </reference>
      <reference anchor="A2A" target="https://a2a-protocol.org">
        <front><title>Agent2Agent (A2A) Protocol</title>
        <author><organization>A2A Project</organization></author><date year="2025"/></front>
      </reference>
      <reference anchor="fullspec" target="https://github.com/mrsergeiluzhnov/xonta/blob/main/spec/ocx/SPEC.md">
        <front><title>Orchestrator Conduct Exchange (OCX): full specification</title>
        <author initials="S." surname="Luzhnov"/><date year="2026"/></front>
      </reference>
    </references>

    <section anchor="ack" numbered="false">
      <name>Acknowledgements</name>
      <t>OCX draws on the models of MCP and A2A for discovery and transport, and
      on Certificate Transparency for its evidence ledger. The design is
      informed by running a two-orchestrator network; the lessons from it are
      recorded in the full specification.</t>
    </section>
  </back>
</rfc>
