Skip to main content
Four Pillars · One Finding Bus · Compound Correlation

The AI Security Layer
Your Apps Need

API Security · MCP Security · AI Gateway · Verification Engine — every pillar publishing to one finding bus.

Most stacks bolt four separate vendors together and lose the thread between them. G8KEPR is one platform with one finding bus — findings raised on the same request are joined by its correlation ID, and separate requests from the same client are joined by organization + client IP.

Multi-Tier Detection
MCP Rug-Pull Detection
Auto Prompt Caching
4-Layer Verification
Append-Only Audit
4
Pillars
1 finding bus
94,000+
Training samples
24 categories
14
LLM providers
wired adapters + BYOI custom
Multi-tier
Detection pipeline
LLM tier off by default
4-layer
Output verification
integrity → constraint
11
Compliance frameworks
auto evidence
2,952
API endpoints
v1 / v2 / v3
21-layer
Middleware pipeline
on every request

4 Products. 1 Platform.

Everything you need to secure APIs and AI applications

API Security

Multi-tier detection · 94,000+ labeled attack samples · OWASP 10/10/10

A multi-tier detection pipeline (regex on every request; an ML classifier on the AI-gateway path; embeddings and NLI when their models load; LLM escalation off by default) with 0.788 recall and 0.857 precision on a 2,580-sample held-out set it was never trained on (3.12% false-positive rate).

ML detection across 24 attack categories (94,000+ labeled attack samples)
Shadow API auto-block · TLS / JA3 fingerprinting
7 evasion vectors handled (base64, NFKC, fragmentation)
Explore API Security

MCP Security

7-step pipeline · default-deny · rug-pull detect

MCP security with SHA-256 rug-pull detection on every tool definition and a default-deny tool-call path.

7-step security pipeline per tools/call
IndirectInjectionScanner · per-tool rate limits · MFA gating
Session replay · cross-session correlation analyzer
Explore MCP Security

AI Gateway

14 providers · auto prompt caching · opt-in adaptive breaker

One API across 14 providers (incl. BYOI custom). Auto-injects Anthropic prompt caching for 88% savings at 10 calls. Adaptive Z-score circuit breaker (opt-in). BYOK with AES-256-GCM.

Multi-factor adaptive routing · 8 guardrail policies
EU AI Act risk-class header on every completion
Hard 16,384 token cap · SSRF-protected transport
Explore AI Gateway

Verification Engine

4 layers · BLOCK-capable · SUGGEST action

Most AI guardrails detect-and-alert after bad output ships. G8KEPR is BLOCK-capable at middleware position 14 — four independent layers, six explicit failure modes, and a SUGGEST action that returns a compliant alternative instead of just an error.

Conversation Integrity · Source Grounding · Tool · Constraint
Two-pass schema enforcement (inject + validate)
Citation grounding · hash-chain drift detection
Explore Verification Engine
MarketplaceComing Soon

Plugin marketplace for extending G8KEPR with community-contributed integrations — not yet shipped. The four pillars above are live today.

One Bus · Compound Correlation

One Finding Bus. All Four Pillars.

Every pillar publishes its findings to one threat event bus. The compound correlator joins findings on the same request by correlation ID, and separate requests from one client by organization + client IP, so a multi-stage attack that crosses pillars becomes one incident.

API Security
WAF · rate limit · ML detector
AI Gateway
14 providers · cache · guardrails
MCP Interceptor
7-step · default-deny · rug-pull
Verification
4 layers · grounding · BLOCK
Audit Chain
SHA-256 hash · 3 verify levels
How Findings Meet On The Bus
User Request → LLM Provider Selected → tools/call Intercepted
↳ Tool Call Checked → Output Verified (4 layers) → Audit Record Written
findings joined on correlation_id (same request) and on organization + client IP (separate requests)
One incident:
A registered multi-stage pattern matched across pillars becomes one compound finding that links every contributing finding as evidence.
Four-vendor stacks can't do this:
Each vendor has its own ID space. Stitching across them after-the-fact is forensic guesswork, not a single query.
What Each Capability Does by Default

Novel Technical Capabilities

Four capabilities in the platform, each marked with what it does on a default install.

Tool Definition Hash Registry

Rug-pull detection. Tool definitions hashed at tools/list time, re-verified on every tools/call. Mid-session mutation blocked with a CRITICAL event.

ENFORCINGSHA-256modules/mcp/tool_registry.py

Adaptive Z-Score Breaker

Statistical baselines per provider per hour-of-day, not static thresholds. Progressive recovery 10/25/50/100%. Off by default; enable with ADAPTIVE_CB_ENABLED.

OPT-IN3σ · 4 windowsgateway/router.py

Cross-Pillar Correlation

All four pillars publish findings to one internal threat event bus, and a correlator scores the ones that occur together into incidents. It joins findings that share a correlation ID (the same request) or come from the same client IP within an organization (separate requests).

ADVISORYone bus · 4 pillarsmodules/compound_correlator/

Append-Only Audit Log

Security events are written to an append-only audit table. Authentication and domain events are HMAC-SHA256 hash-chained per tenant (full, single, last-N verification) and verified daily since 2026-09-13. Mapped to SOC 2 CC7.2, HIPAA §164.312(b), FedRAMP AU-9 (a mapping, not an attestation).

ADVISORYaudit_logs + triggerHMAC-SHA256 chain (auth + domain events)

Platform FAQs

Common questions about the G8KEPR platform

G8KEPR runs alongside the API gateway you already use: your gateway handles routing, G8KEPR handles AI-specific threats and output verification. It brings an ML detector trained on 94,000+ labeled attack samples plus ~340 live regex rules, MCP security with SHA-256 rug-pull detection and a default-deny tool-call path, multi-LLM routing across 14 providers, and 4-layer output verification with a BLOCK-capable SUGGEST action.

Need help choosing the right features?

Talk to our solutions team →
Append-Only Audit Log · 2,000+ Mapped Controls · 14 Frameworks

Auditors Get Exports, Not Spreadsheets

Every pillar writes per-request evidence to an append-only audit log. Control mappings are pre-built for the 14 frameworks listed above — a broader set than the 12 with assessment engines (which score measurable controls from platform evidence). Subject to independent audit and attestation — G8KEPR provides the technical controls and evidence; your auditor issues the certification.

115 controls
NIST 800-53 Rev5
55 controls
PCI DSS v4
197
CSA CCM v4
153
CIS Controls v8
110+
CMMC 2.0
106
NIST CSF 2.0
93
ISO 27001:2022
84
FedRAMP
72
NIST AI RMF
64
SOC 2
48
HIPAA
27
ISO 42001
23
EU AI Act
22
MITRE ATLAS
4 Products in 1 Platform

Start Securing Your APIs
And AI Agents Today

Four pillars, one finding bus, hash-chain audit, and 2,000+ mapped controls across 14 compliance frameworks — without changing a line of your application code.

API + MCP security
14 LLM providers
4-layer verification
Cross-pillar compound correlation

No credit card required • 30-day free trial (extend to 90 days free) • Cancel anytime