Stop Architectural
Derailment.
Fishplate is the first System-Level ORM. It bolts your high-level compliance intent to your low-level implementation, ensuring your AI agents never drift from the Golden Path.
# Requirement: Data Sovereignty
WHEN user_claims == 'EU'
SHALL route_to 'fra-1'
ASSERT encrypted: AES-256-GCM
_
Target: api.segment.io (NIST.SC-7 Violation)
WHAT_IS_A_FISHPLATE
The structural joint for your stack.
Railway fishplates bolt rails into one continuous line. Fishplate does the same for architecture: high-level intent stays bolted to what actually ships.
Alignment
Keeps your API contracts and infrastructure bindings in perfect sync—no silent divergence.
Continuity
Smooth transitions during migrations; one continuous rail from intent to deployed state.
Strength
Braces your system against compliance breaches—NIST, GDPR, and policy as structural load.
SYSTEM_CAPABILITIES
The meta stack for AI-native systems.
High-fidelity controls for contracts, compliance, and runtime truth—hover icons for technical tooltips.
MCP-Native
Built for Cursor and Claude. The AI is your architect, not just your coder.
LAPIS Contracts
85% more token-efficient than OpenAPI—rigorous contracts sized for LLM context windows.
OSCAL Evidence
One-click audit bundles for NIST 800-53, GDPR hooks, and SOC2-oriented evidence.
Native runtime
Low-latency drift detection with optimized parsing of live system state.
EARS Requirements
Structured natural language that bridges human intent and machine-verifiable behavior.
Drift Detection
Continuous reconciliation between your One True Schema and the codebase that runs.
ONE_TRUE_SCHEMA
The closed loop.
Three commands bolt intent to reality. Click a step to focus; read the full guide in documentation.
$ fishplate init → wrote .fishplate/state.json → wrote .fishplate/system.lapis OK twin synchronized
LIVE_CONSOLE
Drift detection, always on.
ENTERPRISE_TIER
Built for the toughest audits.
Move from point-in-time compliance to Continuous Control Monitoring (CCM)—every migration and verification run leaves machine-readable evidence on the rails.
Controls-oriented evidence and drift gates
Data sovereignty and processing constraints as code
Continuous control monitoring, not point-in-time theater