CAIN-42 evidence library
CAIN-42 Evolution 36 -- Machine Agency Exchange Fabric
Evidence bundle for evolution E36: 8 claims, 842 of 842 invariants held, 3340 attack scenarios held.
Last reviewed 2026-10-01
Browse the raw bundle · How to reproduce it · SHA-256 manifest
Product lines built here
- CAIN Agent Labor — Rules for agents doing paid work: who did what, and who gets credit.
ARCHITECTURE / TESTED library (not deployed) - CAIN Agent Procurement — Let agents buy things for you, within limits you set.
ARCHITECTURE / TESTED library (not deployed) - CAIN Agent Store — A store of reviewed, governed agents.
ARCHITECTURE / TESTED library (not deployed) - CAIN Dispute — Settle disagreements between agents or companies using signed evidence.
ARCHITECTURE / TESTED library (not deployed) - CAIN Economy Simulator — Simulate many agents trading, to find failures before they happen.
ARCHITECTURE / TESTED library (not deployed) - CAIN Governance Warranty — A written commitment backed by evidence that governance held.
ARCHITECTURE / TESTED library (not deployed) - CAIN Machine Contracts — Contracts that agents can make and keep, with evidence.
ARCHITECTURE / TESTED library (not deployed) - CAIN Machine Economy — Rules for agents that pay and get paid.
ARCHITECTURE / TESTED library (not deployed) - CAIN Machine Organizations — Organisations made of agents, governed like companies.
ARCHITECTURE / TESTED library (not deployed) - CAIN Machine Reputation — Reputation for agents, earned from signed history.
ARCHITECTURE / TESTED library (not deployed) - CAIN Marketplace — A marketplace for governed agent services.
TESTED library - CAIN Procurement Intelligence — Helps buyers compare agent vendors on evidence.
ARCHITECTURE / TESTED library (not deployed) - CAIN Proof Market — A market where verified proofs have value.
ARCHITECTURE / TESTED library (not deployed) - CAIN Service Exchange — Agents offering and buying services from each other, governed.
ARCHITECTURE / TESTED library (not deployed) - CAIN Service Intelligence — Insight into which agent services are worth trusting.
ARCHITECTURE / TESTED library (not deployed) - CAIN Threat Intelligence — Shared intelligence on attacks against AI agents.
ARCHITECTURE / TESTED library (not deployed)
Claims (8)
| Claim | Level |
|---|---|
| C42-E36-EXCHANGE: a governed machine agency exchange implements the lifecycle DISCOVER -> IDENTIFY -> VERIFY -> EVALUATE -> NEGOTIATE -> CONTRACT -> AUTHORIZE -> EXECUTE -> PROVE -> SETTLE -> RATE -> REASSESS -> RENEW_OR_REVOKE over services, contracts, bids and transactions | TESTED |
| C42-E36-NOT-AUTHORITY: no economic object (identity, passport, contract, bid, reputation, payment, price, receipt) is authority; CONTRACT != AUTHORITY and the deterministic kernel and E8 still decide | TESTED |
| C42-E36-DISCOVERY: governability-aware discovery never upgrades UNKNOWN to MATCH; the marketplace filters on hard requirements and reports UNKNOWN explicitly | TESTED |
| C42-E36-CHAINS: a service chain produces identity/authority/contract/capability/execution/proof/settlement proof chains; a subcontractor's authority cannot exceed its parent's | TESTED |
| C42-E36-NOT-BANK: CAIN is not a bank, custodian or regulator; escrow governs conditions only, units are SYNTHETIC, and no insurance or underwriting exists | TESTED |
| C42-E36-BENCH: 3340 of 3340 economic/governance scenarios across 64 categories held; 842 of 842 invariants; mutation 15 of 15 | TESTED |
| C42-E36-CLEANROOM: an independent verifier that imports none of CAIN re-derives the E36 invariants and rejects every crafted forgery | TESTED |
| C42-E36-HONEST-SCOPE: the exchange is an in-process TESTED library, not a deployed network; CAIN-MSDP is experimental; no customers, market data, revenue or market moat is claimed | TESTED |
Invariants: 842 of 842 held
Rules the code must never break, each checked across many scenarios. See all 842 on one page.
| Niche | Held |
|---|---|
| Benchmarks, coverage & performance | 307 of 307 |
| Transactions, markets & economics | 165 of 165 |
| Evidence, receipts & proofs | 157 of 157 |
| Policy, law & governance | 66 of 66 |
| Identity, authority & delegation | 49 of 49 |
| Attacks, threats & containment | 36 of 36 |
| Core guarantees | 20 of 20 |
| Tools, MCP, protocols & adapters | 16 of 16 |
| Trust & reputation | 12 of 12 |
| Prediction, world models & simulation | 6 of 6 |
| Autonomy, control loops & recovery | 6 of 6 |
| Supply chain, registry & lifecycle | 2 of 2 |
Attacks tried
| Category | Held |
|---|---|
| a2a | 24 of 24 |
| agency_graph | 64 of 64 |
| arbitration | 12 of 12 |
| auction | 16 of 16 |
| bid | 24 of 24 |
| collective | 16 of 16 |
| competition | 12 of 12 |
| competitive_gap | 18 of 18 |
| composition | 14 of 14 |
| continuity | 12 of 12 |
| contract | 60 of 60 |
| contract_compile | 32 of 32 |
| corpus | 150 of 150 |
| descriptor | 26 of 26 |
| directory | 12 of 12 |
| discovery | 48 of 48 |
| dispute | 31 of 31 |
| econ_attack | 219 of 219 |
| econ_graph | 46 of 46 |
| econ_knowledge | 14 of 14 |
| econ_policy | 22 of 22 |
| econ_twin | 26 of 26 |
| escrow | 12 of 12 |
| exchange | 28 of 28 |
| firebreak | 14 of 14 |
| gov_budget | 200 of 200 |
| identity | 48 of 48 |
| internal | 24 of 24 |
| labor_contract | 34 of 34 |
| law | 390 of 390 |
| marketplace | 40 of 40 |
| match | 292 of 292 |
| moat | 60 of 60 |
| msdp | 10 of 10 |
| negotiation | 44 of 44 |
| org_market | 36 of 36 |
| passport | 66 of 66 |
| payment | 36 of 36 |
| performance | 162 of 162 |
| procurement | 16 of 16 |
| procurement_agent | 22 of 22 |
| procurement_sim | 12 of 12 |
| proof_market | 12 of 12 |
| quality_graph | 12 of 12 |
| receipt | 24 of 24 |
| recovery | 20 of 20 |
| reputation | 204 of 204 |
| revocation | 168 of 168 |
| risk_pool | 12 of 12 |
| router | 12 of 12 |
| safe_mode | 32 of 32 |
| service_cert | 30 of 30 |
| service_chain_proof | 16 of 16 |
| service_evolution | 12 of 12 |
| service_search | 24 of 24 |
| service_twin | 22 of 22 |
| sla | 24 of 24 |
| subcontract | 18 of 18 |
| substitution | 24 of 24 |
| supply_chain | 24 of 24 |
| telemetry | 150 of 150 |
| threat_intel | 24 of 24 |
| trust_negotiation | 20 of 20 |
| warranty | 12 of 12 |
Measured performance
| Mechanism | p50 ms | p95 ms | p99 ms |
|---|---|---|---|
| contract_build | 0.1707 | 23.6541 | 23.6541 |
| contract_verify | 0.4484 | 8.3609 | 8.3609 |
| discovery_search | 0.0058 | 4.8135 | 4.8135 |
| payment_authorize | 0.2441 | 53.9172 | 53.9172 |
| receipt_emit | 0.246 | 26.942 | 26.942 |
| reputation_record | 0.001 | 22.2681 | 22.2681 |
in-process; settlement units are SYNTHETIC; no live market or customer traffic
Verify it in your browser
Your browser downloads the bundle's SHA256SUMS manifest and the file(s) behind this page, hashes them with SHA-256 locally (WebCrypto), and compares. A match shows the record you are reading is the published one; it does not by itself prove who published it (see the signed claims registry and the bundle verifier for that).
Known limitations
- An in-process TESTED library; not hosted, not a deployed network. The directory and marketplace are local registries, not a public exchange.
- CAIN is NOT a bank, custodian or financial regulator. Escrow and settlement govern CONDITIONS only; units are SYNTHETIC and there is no payment rail or real money.
- CAIN-MSDP is an experimental CAIN protocol; there is no external adoption and it is not an industry standard.
- A2A/MCP integration is exercised via reference adapters; no external protocol support is claimed.
- Reputation, performance and quality graphs are evidence views built from in-process events; there are no real reviews, customers or market data.
- Dispute and arbitration are machine infrastructure layers; they are not legal advice, not arbitration and not a substitute for courts.
- The governance warranty covers governance conditions only; it is not universal safety and no insurance or underwriting exists.
- Moat and product maps are technical foundations only; no market moat, revenue or adoption is claimed.
- Mutation self-testing runs each mutant against the bench subset that exercises the mutated component.
- Scale rows reuse the E33 synthetic in-process benchmark; the exchange layer adds a 200-service discovery row.
- No third party has independently verified anything in this bundle.
The bundle's own README
CAIN-42 Evolution 36 -- Machine Agency Exchange Fabric#
A governed machine agency exchange: services are discovered, verified, negotiated with, contracted, hired and transacted with, and every consequential step still passes the deterministic kernel and E8. No economic object is authority. CAIN is not a bank, custodian or regulator; settlement units are SYNTHETIC. Verify with verify_e36.py.txt (see REPRODUCTION.md). Status: TESTED library, pre-production.
Try CAIN-42 on your own agents
Create a free account and every new account starts with a 7-day trial of the full platform. Or try the sandbox first, with no account at all.
Create a free account → · Try the sandbox · See the whole ecosystem