CMCI-API helps detect when AI-assisted or dynamic systems begin drifting outside a coherent operating window.
Not a wrapper. Not keyword matching. Structural observability for drift, instability, and loss of coherence.
Coherix Inc. is a Montréal-based (Québec, Canada) R&D company that develops CMCI — Cross-layer Multiscale Coherence Intelligence, a multi-scale structural coherence measurement framework for AI systems and information streams. The framework is peer-reviewed and published in Frontiers in Artificial Intelligence (Vol. 9, 21 May 2026): Dynamic coherence windows: a multi-scale structural coherence measurement framework for AI systems, DOI 10.3389/frai.2026.1836120.
Coherix operates a research-preview dashboard (Coh) and a CMCI Signal API in controlled access. Positioning: AI reliability and observability via structural coherence — not cybersecurity, not cloud infrastructure, not 3D machine vision.
Not affiliated with Coherix, Inc. of Ann Arbor, Michigan, USA — a separate company in 3D machine vision and adhesive-process control.
Coherix engages via pilots, partnerships, and controlled licenses, not as a self-serve consumer product.
CMCI sits inside a wider research program — theory, methodology, applications, partners, publications. The Coherix Atlas surfaces all eight domains and how they connect.
Static archival graph view also available at layered observatory.
The same coherence mathematics that runs in software is being explored as a hardware architecture. A Verilog HDL design of a multiscale-coherence core has been written and validated in simulation (Verilator + a bit-exact Python verifier). No FPGA has yet been programmed and no silicon has been taped out. Hardware acceleration is a R&D direction, not a current product.
The proposed core implements five parallel Hardware Coherence Units (HCU) feeding an aggregator and a coherence gate. The HDL has been simulated; no real FPGA or ASIC is currently in production.
cmci_core.v,
structural_aggregator.v, gate_block.v, etc.The dev-board path lets us validate the CMCI hardware pipeline before committing to a custom silicon spin. It preserves capital and proves the math at silicon speed.
CMCI-FPGA is the first step toward a broader class of coherence-aware computing substrates: motherboards and accelerators that continuously monitor their own structural state, not just their throughput.
CMCI measures structural integrity across every level of organization. When coherence degrades at any scale, CMCI detects it — before it becomes critical.
CMCI tracks five fundamental quantities that govern the structural health of any information stream. These aren't sentiment scores or keyword counts — they're dynamical state variables from complex systems theory.
CMCI doesn't just measure overall quality. It decomposes structure into four distinct scales, then measures how they interact.
Per-observation level. Local coherence of each incoming signal: stability, continuity, regularity. The building blocks of the state trajectory.
Regime level. Structural organization of the session into identifiable coherence regimes, and the transitions between them.
Full-session level. Global transition economy, long-range coherence, and whether the architecture as a whole remains viable.
Cross-scale compatibility. Do micro, meso and macro remain mutually coherent? This is where CMCI is distinctive.
CMCI identifies five distinct failure modes. Each axis is a latent state variable that measures a specific type of structural fragility in multi-agent pipelines.
Coherix was born from a fundamental question: can structural coherence be measured mathematically? Founded in Montréal, the CMCI framework emerged from dynamical systems theory — applying the mathematics of phase transitions and critical phenomena to information structure. The framework is published in Frontiers in Artificial Intelligence (Vol. 9, 21 May 2026): Dynamic coherence windows: a multi-scale structural coherence measurement framework for AI systems, DOI 10.3389/frai.2026.1836120. Intellectual property is protected (patent pending; internal reference OPIC #3304899).
The central innovation of CMCI. A coherence window defines the viability space where a system maintains structural integrity. When defects push the system outside this window, coherence rupture occurs.
The coherence window W is defined as the set of all defect configurations where every defect remains below its critical threshold: W = {D | ∀i: di < δi}. The coherence margin m = mini(δi − di) measures the distance to rupture. When m approaches zero, the system is about to exit the window — CMCI detects this before it happens.
CMCI doesn't just detect failure — it predicts how it will happen. Transversal resonance analysis tracks oscillations between defects to identify whether energy is accumulating or dispersing.
Traditional analysis tells you something is wrong. CMCI tells you what kind of wrong, where it's heading, and how fast. The transversal resonance module tracks a 5D velocity vector through defect space and detects oscillation patterns between defect pairs.
When defects oscillate in-phase, energy accumulates — a phase transition is imminent. When they oscillate out-of-phase, energy disperses — the system self-protects. This is the difference between knowing "coherence is dropping" and knowing "the system is about to fragment via scale separation."
Defects in-phase. Energy accumulates. Phase transition imminent.
Defects out-of-phase. Energy disperses. System self-protects.
Mode switching. Resonance pattern is changing — critical observation point.
Insufficient data. Observation window still accumulating measurements.
The Cognitive Immune System (CIS) is a preventive evaluation pipeline introduced in the Frontiers paper. It evaluates each new input and routes it through one of four decision pathways:
Active Coherence Recovery is the direction that materializes the full loop — observe → measure → detect → diagnose → guide → intervene → re-measure → verify — in five maturity levels:
Published limitation (verbatim from the Frontiers manuscript): The CIS has been implemented and tested in controlled scenarios but has not yet been evaluated in large-scale production deployments. The effectiveness of human coherence bridging depends on operator expertise.
Source: Christian St-Louis, Dynamic coherence windows: a multi-scale structural coherence measurement framework for AI systems, Frontiers in Artificial Intelligence, Vol. 9, 21 May 2026 — DOI 10.3389/frai.2026.1836120.
GET /health returns 200,
POST /diagnose-structure returns 200 with
privacy_mode=content_not_received and
diagnosis_status=not_wired; restricted routes return 404.One product, empirically validated. Real-time structural coherence monitoring for multi-agent AI pipelines — push observations, get back regime, safety margin, and five defect axes with calibrated per-axis thresholds.
REST endpoints to observe, monitor, and explain multi-agent drift in real time. Per-axis empirical thresholds, deterministic fallback, validator-gated LLM narration.
REST · Pilot keysWatch CMCI detect drift → rupture → recovery on a scripted multi-agent scenario. Real engine, real thresholds, real narration.
No signup · Coherix sessionCMCI serves two kinds of buyers: researchers whose work is episodic, and organisations whose production AI systems require ongoing structural coherence. A single tier ladder served neither well — here is the new structure. All prices in USD. Version française →
Time-boxed. No auto-renewal. Designed for the episodic rhythm of research work.
For researchers affiliated with an institution.
One-time · no auto-renewal · institutional email + 1-page proposal required
Independent researchers, R&D teams, OSS maintainers.
One-time · no auto-renewal · 90-day extension available for $99
Transition clause: If you move to a Commercial tier within 60 days of pass expiry, the pass fee is credited against your first invoice.
Platform and services priced separately. Services are where the differentiation lives.
Validate fit on one system before committing.
Flat fee · 3 weeks · payable at signing
Credited if you sign Core/Enterprise within 30 days
Continuous monitoring with calibrated thresholds.
Platform fee · monthly
+ $1,500 setup engagement (one-time)
+ $750 per quarterly review
Annual prepay: $5,500 (≈ 20% discount)
Multi-system governance with SLA.
Platform · custom range
+ $5k–15k initial discovery
+ $3k–8k quarterly retainer
Embed CMCI as a differentiating component.
Annual minimum from $50,000
+ Royalty on derived revenue
The "so what / now what" layer that translates CMCI diagnostics into actionable decisions. Four to six pages, executive-readable, fixed four-section structure:
Get instant insights into structural integrity across every scale.