# CAIN-42 — Canonical Definition (2026-11-13)

**Status: canonical. Supersedes earlier one-line descriptions that called CAIN-42 only a security tool.**

## The definition

**CAIN-42 is a Byzantine trust fabric and the governed execution paradigm for autonomous AI.**

It is the runtime trust and control layer for the autonomous economy: the system that establishes who an agent is, what authority it holds, what it intends, what it is permitted to do, what actually happened, and whether it may continue — at the exact boundary where a machine action becomes a real-world effect.

CAIN-42 is **not** a chatbot, a dashboard, an API gateway, an MCP firewall, policy software, or an AI security scanner. Those are surfaces it can expose. The system itself is larger:

- a **trust kernel** (16-stage causal execution, capability tokens, temporal trust, trajectory continuity),
- a **Byzantine evidence fabric** (4-replica PBFT quorum, Ed25519, SHA-256 chains, RFC 6962 Merkle roots, offline clean-room verifiers),
- a **governed mission runtime** (AOS: missions, budgets, child grants, recovery, skills),
- a **distributed workforce fabric** (E68: real worker processes, ticket-bound authority, cross-field independent verification),
- an **identity and delegation fabric** (E61/E62: ownership, workload identity, attenuation, revocation),
- a **memory authority layer** (AETHER-33: memory is never permission),
- an **isolation and attestation fabric** (sandbox, workload attestation appraisal, Byzantine settlement),
- and a **product factory** — the **Agentic 75** — that packages the whole system into fixed-scope deliverables.

## The Agentic 75 — the operating model for agentic AI in 2027

The Agentic 75 maps the 75 defining agentic-AI developments of 2026 to the seven moats and packages each one as a governed product with:

- a published rule set (`rules.py`),
- a runnable engine (`engine.py`),
- a signed sample and a standalone offline verifier,
- a public product page and a live try-it route,
- an explicit status and an explicit limit.

It is the shift from **"AI that can act"** to **"AI that can prove it was allowed to act."** That is the paradigm: not better prompting, not another observability pane — a governed execution boundary with evidence that survives the vendor.

The doctrine in one line:

> **AGENT PROPOSES → CAIN DECIDES → MCPGATE ENFORCES → WORKER EXECUTES → INDEPENDENT AGENT VERIFIES → CLAWX PRESERVES.**

## What is real today (measured, not asserted)

| Surface | Evidence |
|---|---|
| Agentic 75 | **75 / 75 DELIVERABLE**, each with a dedicated engine package, published rules, signed sample and offline verifier |
| Product tests | **2,598** Agentic 75 tests pass; catalog selftest **75/75** |
| Trust kernel | 16-stage causal execution context; 26 machine-checkable kernel checks |
| Consensus | Live PBFT cluster `cain-mr-02`: N=4, f=1, Q=3, four regions, vendor-operated |
| Evidence | v5 mission bundles; clean-room verifier imports no CAIN code; pinned-key trust model |
| Workforce | **1,000 real worker processes**, 2,000 independently verified tasks; cross-field verification |
| Continuous assurance | ARGUS hosted property suite **117/117**; restriction-only stages in the hosted path |
| Public surface | `cainstudio.online`, `mcpgate.online` (+ subdomains), `clawx.click`, Hugging Face Space, machine-readable `llms.txt` |

## What it is not (stated without minimizing the system)

- It is not certified by a third party, and it has no SOC 2, ISO 42001, or EU AI Act certification.
- It has no patents filed.
- It has one paying subscriber (live Stripe) and a small, founder-attested customer base.
- Some engines are deterministic and tested; some adjacent paths remain prototypes or simulations, and each product page says which.
- Consensus notarizes evidence; it never authorizes an action.
- The quorum is operated by FIRME AI CORPORATION, not an independent witness network.

Those are boundaries on the evidence, not a reduction of the architecture. The architecture is a trust fabric for autonomous systems. The Agentic 75 is the operating model that makes it usable, one governed product at a time.

## Canonical links

- Everything: https://cainstudio.online/cain-42.html
- 75 products: https://cainstudio.online/agentic-products
- 75 governance pages: https://cainstudio.online/agentic-75
- Catalog JSON: https://cainstudio.online/agentic-products.json
- Signed evidence: https://clawx.click/evidence/agentic-75-products-2026-11-12/
- Hugging Face Space: https://huggingface.co/spaces/vicentenoriega/CAIN-42
- Machine-readable definition: https://cainstudio.online/llms.txt

**FIRME AI CORPORATION**, California. Founder: Vicente Fernando Noriega. Team: Vicente Fernando Noriega, co-founder, co-founder.
