INVARIANS - Nominal Windows for AI Agents on Blockchain. Safer Transactions for RWA and Institutional Flows.

AI agents now execute on-chain at scale. The blockchain substrate they run on is not nominal at all times. Yet most agents act as if it were.

Invarians is the on-chain execution context that tells an AI agent when conditions align to act, and signs that judgment.

Who it is for

• RWA agents, tokenized treasuries, real-estate share registries, settlement orchestrators.
• Institutional flows, treasury management, custody, programmatic rebalancing.
• Compliance-bound flows, audit-trail SLAs, regulator-facing reporting.
• Cross-chain orchestration, CCIP, CCTP, native canonical bridges.
• Audit-grade execution, a cryptographically signed proof of the infrastructure conditions under which the agent acted.

How

Two operations, both calibrated per chain on documented events:

1. Target nominal windows. The substrate operates within a measurable zone of low variance, derived from years of historical data and bound by ground-truth incidents (The Merge, Solana outages, Polygon Reorg Storm, OP sequencer events). Nominal is not assumed; it is the densest sample of the distribution.
2. Characterize divergences through calibration. Every threshold has a published TPR and FPR with Clopper-Pearson 95 percent confidence intervals. Every recalibration is logged. Methodology is open at agentnorthstar.com/calibration.

Three primitives, one signed payload

API v2.0 unifies three explicit primitives in a single response, signed HMAC-SHA256 and verifiable independently:

1. Attestation. Cryptographic envelope. Answers: can I trust this data?

2. Regime. Substrate state classification per layer. Twelve signed codes per chain on the calibrated cohort, plus bridge state. Answers: what is the substrate doing now?

3. Drift Signal. Continuous reading of how far each classifying metric has moved from its long-term baseline, and whether it is amplifying or reverting. Per metric: shift, shift_delta,
   shift_magnitude_delta. Answers: is it safe to act in the next 30 minutes?

The matrix

(Sx±.Dx±).Lx.BSx

• S = Structural metrics (rhythm, continuity, beacon participation, sequencer publish latency)
• D = Demand echo (sigma, size, transactions, composition)
• x = 1 (nominal) or 2 (stress)
• ± = sub-nominal or supra-nominal (bilateral calibration)
• L = Layer (L1, L2, bridge)
• BS = Bridge structural state (BS1 nominal, BS2 degraded)

A composite execution context per cross-layer action. Direction is composed by the agent from the panel, not prescribed by Invarians.

What is being researched

Blockchains as materials under agentic load. Invarians Labs measures the deformation of the substrate as AI agents (bundlers, MEV builders, keepers, intent solvers, x402 micro-payments) raise the load. Two estimators:

• σ, the agentic pressure, a composite of seven on-chain footprints normalized on a 30-day rolling z-score.
• ε, the substrate deformation, the universal four-ratio composite uniformly defined on L1 and L2.

Three rheological regimes are tested against the data: elastic (substrate returns to nominal when load drops), plastic (the nominal itself shifts), and creep (slow accumulation under sustained sub-yield load). Rupture is not claimed. The baseline started in March 2026 and cannot be reconstructed retroactively.

Public commitments

• API v2.0 live since 2026-04-30.
• Calibration methodology, backtests, scripts published at agentnorthstar.com/calibration
• Honest gaps documented at the same address.
• Invarians signs what the substrate is doing. The agent owns the execution policy.

Concrete: pip install invarians returns the panel in under 100 ms, signed and verifiable. Documentation at invarians.com/developers.