Documentation / Getting Started

Getting Started with EVE AI Core

Integrate deterministic governance enforcement into your AI pipeline. Govern every proposed action before it executes — and, optionally, every output before it ships.

What EVE AI Core Does

EVE AI Core is a deterministic governance enforcement platform. It sits at the execution boundary — between your AI system and the action it is about to take — evaluating each proposed action (a tool call, MCP request, response, or multi-step action sequence) against configurable charter rules before it executes.

Every proposed action routed through EVE is evaluated against a configurable set of charter rules, scored using Calibrated Reliability Diagnostics (CRD), and returned with a deterministic verdict — ALLOW, BLOCK, or MODIFY — before any side effect occurs. The platform provides cryptographic audit trails, circuit breaker protection against provider failures, and a truth store for grounding factual claims.

Because the decision is made before execution, a BLOCK prevents the side effect rather than reporting it after the fact, and a MODIFY returns a policy-constrained alternative in place of the original. Every verdict ships with a structured, signed explanation of what was decided and why. Governing model responses after generation (before they reach a user) is available as an optional post-generation layer on the same engine.

Core principle: Governance decisions are deterministic. The same input and rule set always produce the same verdict. No probabilistic filtering, no hidden model-based overrides.

How to Get Started

1

View the Live Demo

See governance enforcement in action. The demo shows real-time CRD scoring, charter rule evaluation, circuit breaker status, and verdict generation against sample AI outputs.

Run the Live Demo View Licensing
2

Review the API Documentation

The API reference covers authentication, endpoint specifications, request/response formats, error codes, and rate limits. All governance endpoints return structured JSON verdicts with full audit metadata.

API Reference Python SDK Guide
3

Install the SDK

Official SDKs are available for Python (eve-coreguard) and Node.js (eve-ai-governance), both with zero heavy dependencies. Any other stack can call the REST endpoint directly.

# Python
pip install eve-coreguard

# Node.js
npm install eve-ai-governance

# Any language: call the REST API directly
POST https://api.eveaicore.com/v1/decisions/evaluate

Quick Integration Example

The following Python example shows a minimal integration. Obtain your API key from the developer console after your license is provisioned. evaluate() takes keyword arguments directly — there is no request object to construct.

from eve_coreguard import CoreGuardClient

# Initialize with your API key
client = CoreGuardClient(api_key="eve_sk_...")

# The action your AI proposed — evaluate it before it is acted on
proposed_action = {"type": "loan_approval", "amount": 250000}

result = client.evaluate(
    tenant_id="bank_001",
    proposed_action=proposed_action,
    model_output={"decision": "approve", "confidence": 0.91},
    context={"credit_score": 580, "debt_to_income": 0.52},
    policy_set="lending_v1",
)

# Act on the verdict: ALLOWED | BLOCKED | MODIFIED
if result.blocked:  # BLOCKED — the action is refused
    # result.policy_violations explains every rule that fired
    use_safe_alternative()
elif result.verdict == "MODIFIED":
    # policy rewrote the action — act on result.modifications, not the original
    proceed(result.modifications)
else:  # ALLOWED
    # result.decision_id + result.signature = the signed certificate
    proceed(proposed_action)
Every decision is a signed certificate. Each evaluation returns a signed Decision Certificate (result.decision_id, result.signature, result.audit_hash) — Ed25519 in production configuration, with an HMAC-SHA256 fallback when asymmetric keys are not configured. Your controls team can verify it offline with the Ed25519 public key, with no call back to EVE (the HMAC fallback is symmetric and not independently verifiable). Governance verification adds single-digit milliseconds to your pipeline.

Key Concepts

CRD Scoring

Calibrated Reliability Diagnostics (CRD) is the composite integrity score assigned to every AI output. CRD evaluates factual grounding, charter compliance, hallucination risk, and uncertainty calibration. Scores range from 0 to 100. Outputs below your configured threshold are automatically flagged or blocked.

Charter Rules

Charter rules are the deterministic governance constraints applied to every output. EVE ships with 15 immutable rules covering safety, identity integrity, non-harm, and truthfulness. These cannot be weakened. Enterprise customers can add custom rules that impose additional restrictions on top of the immutable set.

Circuit Breaker

The circuit breaker monitors upstream AI provider health. When a provider begins failing (consecutive errors, latency spikes, rate limit exhaustion), the circuit breaker opens and routes traffic to fallback providers or returns a structured degradation response. This prevents cascading failures in your production pipeline.

Truth Store

The truth store is a curated, versioned repository of verified facts used to ground AI outputs. When an output makes a factual claim, the governance pipeline checks it against the truth store. Claims that contradict verified facts are flagged with specific correction metadata. Enterprise customers can populate the truth store with domain-specific ground truths.

CRD Scoring

Composite integrity score (0-100) for every verified output

Charter Rules

15 immutable rules + custom enterprise rules

Circuit Breaker

Provider health monitoring with automatic failover

Truth Store

Verified fact repository for grounding AI claims

Audit Trail

Cryptographic hash-chained logs for every verdict

Attestation

Ed25519-signed certificates for compliance

Enterprise Licensing

EVE AI Core is licensed on an annual basis. Most engagements start with a fixed-fee Design Partner Pilot — proving enforcement on your own decision logs — then convert to a production license. See the full pricing page for current details.

Design Partner Pilot

$37,500 (60-day)

Sandbox pilot that proves enforcement on your own decision logs — single tenant, one policy pack, no production data required. The pilot fee is credited toward your first annual license.

CoreGuard Enterprise

From $150,000/year

Production enforcement for a regulated workflow. $150K/year covers up to 500K governed decisions per month, with higher tiers as volume grows; signed decision records for every governed action, 3-year retention with open export, and a 99.5% uptime SLA with a 24/7 sev-1 path.

Enterprise Governance

Contact Sales

Multi-workflow governance across teams with tenant isolation. Custom decision volumes, 7-year retention, on-premise deployment option, and a 99.9% uptime SLA.

Sovereign Infrastructure

Contact Sales

For Tier-1 banks, large insurers, and defense. Any deployment including air-gapped, unlimited governed decisions, white-label attestation, and a custom negotiated SLA.

Assessment first: All engagements begin with a governance assessment to scope integration requirements, compliance needs, and deployment architecture. Contact the governance team to schedule.
Schedule Assessment View Full Licensing Details

Next Steps

Live Demo

See governance enforcement in action

API Reference

Full endpoint documentation

Licensing

Tier comparison and engagement process

Patent Portfolio

93 filed applications covering governance IP

FAQ

Common questions about the platform

Contact

For governance licensing inquiries, integration support, or to schedule a platform assessment:

Part of the EVE AI Core control plane Deterministic AI Governance Control Plane → Policy decisions that return the same result for the same input every time, before execution.