Skip to content

HunterSpence/enterprise-ai-accelerator

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

32 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Enterprise AI Accelerator

AI-native unified cloud governance platform — multi-cloud discovery, 6R migration planning, IaC security, FinOps intelligence, compliance audit, and executive AI chat. Built entirely on Claude Opus 4.7. Zero paid SaaS dependencies.

Tests Python 3.11+ License: MIT EU AI Act FOCUS 1.3 Claude Opus 4.7 Prompt Caching Extended Thinking 1M Context Batch API IaC Security Multi-Cloud OpenTelemetry Carbon Aware

April 2026 — Opus 4.7 Executive Upgrade + v0.2.0 Platform Expansion. Platform now runs on Claude Opus 4.7 across every auditable path, with prompt caching, native tool-use structured output, extended-thinking reasoning traces as Annex IV evidence, a 1M-context executive chat, Batch API bulk scoring, and seven new capability tracks (multi-cloud discovery, IaC security, app portfolio scanning, integration hub, observability, advanced FinOps, and an Anthropic-native cost optimization layer). See docs/OPUS_4_7_UPGRADE.md and CHANGELOG.md for details.


What this is

Enterprise AI Accelerator is an AI-native unified cloud governance platform built exclusively on Claude Opus 4.7 and open-source dependencies. It replaces the fragmented point solutions — migration tools, IaC scanners, FinOps dashboards, compliance auditors — that enterprise teams currently assemble from five to ten separate vendors, and does so at a fraction of the cost with a single audit trail. The platform covers the full cloud governance lifecycle: discover your multi-cloud estate, classify workloads for migration, scan infrastructure code for security and compliance violations, optimize cloud spend down to carbon emissions, and surface every decision in a tamper-evident audit chain that satisfies EU AI Act Annex IV. Everything runs on a single Anthropic subscription with no paid SaaS intermediaries.


What This Replaces

Bottom line. A Big 6 cloud-transformation engagement delivers a PowerPoint in 12–18 months for $3.2M–$12M. This platform ships equivalent technical surface area today for $25K–$80K / year — one Anthropic subscription, zero paid SaaS, a single tamper-evident audit trail.

Big 6 consulting platforms

Firm / Platform Engagement What they ship What this platform ships instead
Accenture — MyNav $500K–$5M 2019-era recommendation engine with GPT-4 wrappers; no persisted reasoning traces Native Opus 4.7 reasoning; extended-thinking traces as Annex IV audit evidence; source-available
Deloitte — CloudCompass / Converge $400K–$3M + $200K/yr Tableau + Excel + PowerPoint automation; human-curated compliance content Agentic decision engine; Citations API grounds every compliance claim in cited regulation text; SARIF 2.1.0 export
PwC — Cloud Transformation Suite $300K–$2M LLMs used only for summarization; assessment-heavy Frontier-model decision engine; tamper-evident Merkle audit chain; integrated carbon tracking
EY — Nexus for Cloud $400K–$2M Thin glue over AWS Migration Hub / Azure Migrate; PowerPoint deliverables Unified 6R + compliance + FinOps on one audit trail; structured tool-use JSON output
KPMG — Powered Enterprise Cloud $500K–$4M SAP-centric; light on cloud-native / K8s / serverless Multi-cloud adapters (AWS / Azure / GCP / K8s); IaC drift detection; open-source carbon model
Cognizant — CloudVue / Cloud Steps $200K–$1.5M Pattern-matching over rules tables Genuine extended-thinking reasoning; prompt caching; Anthropic-native architecture
Capgemini — eAPM + Migration Factory $300K–$2M Rules-based app portfolio analysis AI-native 6R scoring; 1M-context executive chat; interleaved thinking + tool-use

Where the Big 6 still win. Brand trust with boards, regulatory sign-off at Fortune 500 scale, and the organizational change-management muscle a 200-person program office provides. This platform is the technical substrate — not a replacement for that program office.

Commercial point tools

Tool Category List price Replaced by
CAST Highlight / vFunction App portfolio + refactoring $150K–$600K/yr app_portfolio/
Snyk IaC / Prisma Cloud IaC security $200K–$1M/yr iac_security/
IBM OpenPages AI Governance AI risk + compliance $500K–$2M/yr ai_audit_trail/ + policy_guard/ + compliance_citations/
Credo AI Bias + AI compliance $180K/yr policy_guard/thinking_audit.py
ServiceNow IRM + AIOps GRC + ops $300K–$2M/yr integrations/ + executive_chat/
Apptio Cloudability / CloudZero FinOps $100K–$1M/yr finops_intelligence/
Flexera One / Turbonomic Cloud cost + optimization $100K–$2M/yr cloud_iq/adapters/ + finops_intelligence/right_sizer.py
Datadog / New Relic Observability $50K–$500K/yr observability/ (OSS OTEL + Prometheus + Grafana)
Cloud Carbon Footprint Sustainability Free finops_intelligence/carbon_tracker.py

3-year TCO — 10,000-workload enterprise

Approach Year 1 Year 3 cumulative
Big 6 engagement + licensed tools $3.2M–$12M $6M–$25M
Best-of-breed commercial tools, assembled in-house $1.5M–$4M $4.5M–$12M
This platform (Anthropic API + self-hosted) $25K–$80K $75K–$240K

Platform diagrams

Platform data flow

graph TD
    A["User / Executive"] --> B["MCP Server (18 tools + 4 resources + 4 prompts, stdio + SSE)"]
    B --> C["AgentOps Orchestrator"]
    C --> D["Opus 4.7 Coordinator"]
    D --> E["ArchitectureAgent (Haiku 4.5)"]
    D --> F["MigrationAgent (Haiku 4.5)"]
    D --> G["ComplianceAgent (Haiku 4.5)"]
    D --> H["ReportAgent (Sonnet 4.6)"]
    E --> I["AI Audit Trail"]
    F --> I
    G --> I
    H --> I
    I --> J["SHA-256 Merkle chain + SARIF 2.1.0"]
Loading

Module coverage

graph TD
    Core["Core AIClient — caching, thinking, routing, streaming, batch, citations"]
    Core --> CloudIQ["cloud_iq/adapters/ — AWS / Azure / GCP / K8s discovery"]
    Core --> Scout["migration_scout/ — 6R classifier + batch + thinking audit"]
    Core --> Portfolio["app_portfolio/ — repo to 6R via extended thinking"]
    Core --> IaC["iac_security/ — Terraform + Pulumi + SBOM + OSV + drift + SARIF"]
    Core --> FinOps["finops_intelligence/ — CUR + RI/SP + right-size + carbon"]
    Core --> Policy["policy_guard/ — IaC + bias + thinking audit"]
    Core --> Citations["compliance_citations/ — Citations API evidence grounding"]
    Core --> Chat["executive_chat/ — 1M-context unified Q&A"]
    Core --> Integrations["integrations/ — Slack / Jira / ServiceNow / Teams / GitHub / PagerDuty"]
    Core --> Obs["observability/ — OTEL + Prometheus + Grafana"]
Loading

Quick Start

git clone https://github.com/HunterSpence/enterprise-ai-accelerator.git
cd enterprise-ai-accelerator
pip install -r requirements.txt
export ANTHROPIC_API_KEY=sk-ant-...

# Simplest demo — scan a local repo for app portfolio intelligence
python -m app_portfolio.cli .

# AI governance + EU AI Act compliance
python -m ai_audit_trail.demo

# Multi-cloud discovery (auto-detects available credentials)
python -c "from cloud_iq.adapters.unified import UnifiedDiscovery; UnifiedDiscovery.auto().discover()"

# IaC security scan
python -m iac_security .

# Full FinOps with CUR ingestion + carbon tracking
python -m finops_intelligence.demo

All module demos include synthetic data. No cloud credentials required to run any demo.


Architecture at a Glance

┌─────────────────────────────────────────────────────────────────────────────────┐
│                              Entry Points                                        │
│       CLI  ·  MCP Server (19 tools)  ·  Python SDK  ·  Webhook Dispatcher       │
└──────┬──────────────────┬────────────────────┬──────────────────────────────────┘
       │                  │                    │
       ▼                  ▼                    ▼
┌─────────────────────────────────────────────────────────────────────────────────┐
│                         core/ — Anthropic Optimization Layer                     │
│  AIClient · ModelRouter (~95% cost savings) · ResultCache · BatchCoalescer      │
│  Streaming · FilesAPI · InterleavedThinking · CostEstimator · Telemetry         │
└──────┬──────────────────┬────────────────────┬──────────────────────────────────┘
       │                  │                    │
┌──────▼──────┐  ┌────────▼────────┐  ┌────────▼───────┐  ┌───────────────────────┐
│  cloud_iq/  │  │  app_portfolio/ │  │  iac_security/ │  │  finops_intelligence/ │
│  adapters/  │  │  (11 languages) │  │  (20 policies) │  │  CUR + RI/SP + right- │
│  AWS·Azure  │  │  OSV CVE scan   │  │  SBOM·SARIF    │  │  sizing + carbon      │
│  GCP·K8s    │  │  6R via Opus    │  │  drift detect  │  │  DuckDB analytics     │
└──────┬──────┘  └────────┬────────┘  └────────┬───────┘  └──────────┬────────────┘
       │                  │                    │                      │
       └──────────────────┴────────────────────┴──────────────────────┘
                                    │
┌───────────────────────────────────▼────────────────────────────────────────────┐
│              agent_ops/ — Multi-Agent Orchestrator                              │
│   Opus 4.7 Coordinator · Sonnet 4.6 Reporter · Haiku 4.5 Workers               │
└───────────────────────────────────┬────────────────────────────────────────────┘
                                    │
          ┌─────────────────────────┼─────────────────────────┐
          │                         │                         │
┌─────────▼──────────┐   ┌──────────▼──────────┐   ┌─────────▼──────────────────┐
│  migration_scout/  │   │   policy_guard/     │   │   ai_audit_trail/           │
│  6R + Monte Carlo  │   │  EU AI Act + HIPAA  │   │   SHA-256 Merkle chain      │
│  dependency maps   │   │  SOC2 + PCI-DSS     │   │   SARIF 2.1.0 + Article 12  │
│  wave planning     │   │  SARIF 2.1.0        │   │   Annex IV evidence         │
└────────────────────┘   └─────────────────────┘   └────────────────────────────┘
          │                         │                         │
          └─────────────────────────▼─────────────────────────┘
                                    │
┌───────────────────────────────────▼────────────────────────────────────────────┐
│              executive_chat/ + compliance_citations/ + risk_aggregator.py       │
│   1M-context CTO Q&A  ·  Citations API compliance evidence  ·  0–100 score     │
└───────────────────────────────────┬────────────────────────────────────────────┘
                                    │
┌───────────────────────────────────▼────────────────────────────────────────────┐
│              integrations/ + observability/                                     │
│   Slack · Jira · ServiceNow · GitHub · Teams · PagerDuty · SMTP                │
│   OTEL gen_ai.* traces · 8 Prometheus metrics · Grafana dashboards             │
└────────────────────────────────────────────────────────────────────────────────┘

Model tier: Opus 4.7 handles coordination + high-stakes reasoning (6R, extended thinking, executive chat). Sonnet 4.6 handles report synthesis. Haiku 4.5 handles high-volume worker tasks. The model router selects the right tier automatically based on task complexity.


Module Reference

Module Purpose Key Classes Value Prop
core/ Anthropic optimization layer AIClient, ModelRouter, ResultCache, BatchCoalescer, CostEstimator, StreamHandler, FilesAPIClient, InterleavedThinkingLoop ~95% cost reduction vs always-Opus baseline via complexity routing + SQLite cache + auto-coalescing Batch API
cloud_iq/ AWS infrastructure analysis CloudScanner, CostAnalyzer, MLDetector, NLQueryEngine $47K/month waste identified in AcmeCorp demo without credentials
cloud_iq/adapters/ Multi-cloud discovery AWSAdapter, AzureAdapter, GCPAdapter, KubernetesAdapter, UnifiedDiscovery Real boto3 / azure-mgmt / google-cloud / kubernetes discovery with graceful degradation
app_portfolio/ Repository intelligence LanguageDetector, DependencyScanner, CVEScanner, ContainerizationScorer, CIMaturityScorer, SixRScorer 11 languages, 9 dep manifests, OSV.dev CVE scan, Opus 4.7 extended-thinking 6R per repo
migration_scout/ 6R workload classification WorkloadAssessor, DependencyMapper, WavePlanner, BatchClassifier, ThinkingAudit Only OSS tool with AI-native 6R + Monte Carlo wave planning (AWS Migration Hub closed Nov 2025)
policy_guard/ Multi-framework compliance ComplianceScanner, BiasDetector, SARIFExporter, IncidentResponse, ThinkingAudit 9-framework scanner (EU AI Act + HIPAA + SOC 2 + CIS AWS + NIST AI RMF 2.0 + ISO 42001 + DORA + FedRAMP Rev 5 + PCI DSS 4.0); cross-framework traceability matrix; one implementation closes controls across all frameworks simultaneously
iac_security/ IaC security scanning TerraformParser, PulumiParser, PolicyEngine, SBOMGenerator, OSVScanner, DriftDetector, SARIFExporter 20 built-in policies (CIS AWS / PCI-DSS / SOC 2 / HIPAA), CycloneDX SBOM, OSV CVE, SARIF to GitHub Security tab
finops_intelligence/ Cloud cost intelligence CURIngestor, RISPOptimizer, RightSizer, CarbonTracker, SavingsReporter, AnomalyDetector AWS CUR via DuckDB, RI/SP optimizer (80% coverage cap), right-sizing with CloudWatch, carbon tracking with open coefficients
ai_audit_trail/ EU AI Act audit logging MerkleChain, EUAIActLogger, NISTRMFScorer, IncidentManager, SARIFExporter Only OSS tool combining SHA-256 Merkle chain + SARIF 2.1.0 + Article 12 / Annex IV
executive_chat/ 1M-context CTO Q&A ExecutiveChat, BriefingLoader Full enterprise briefing in one prompt; follow-ups cost ~10% via 1-hour cache
compliance_citations/ Evidence-grounded compliance EvidenceLibrary, CitationsEngine Anthropic Citations API — character-range citations, no hallucinated control IDs
agent_ops/ Multi-agent orchestration Orchestrator, CoordinatorAgent, ReporterAgent, WorkerAgent Opus 4.7 coordinator + Sonnet 4.6 reporter + Haiku 4.5 workers with MCP-driven dispatch
integrations/ Notification + ticketing FindingRouter, WebhookDispatcher, SlackAdapter, JiraAdapter, ServiceNowAdapter, GitHubAppAdapter, TeamsAdapter, PagerDutyAdapter, SMTPAdapter Retry / circuit-breaker / rate-limit on all adapters; PR check-runs with inline annotations
observability/ Full OTEL stack TelemetryClient, PrometheusExporter, Grafana dashboards gen_ai.* conventions, 8 Prometheus metrics, Grafana platform + cost dashboards, Jaeger traces
risk_aggregator.py Cross-module risk score WorkloadRiskAggregator, RiskInput Unified 0–100 score from any combination of module outputs
mcp_server.py MCP surface 18 tools + 4 resources + 4 prompts Every module drivable from Claude Code / Claude Desktop; stdio + SSE transports

Capabilities by Theme

Theme What the platform covers
Discovery Real boto3/azure-mgmt/google-cloud/kubernetes discovery; 11 programming languages; 9 dependency manifest formats; OSV.dev CVE feed
Migration Planning AI-native 6R classification; Monte Carlo wave planning with confidence intervals; dependency SCC resolution; 3-year TCO; AWS MAP alignment
Compliance EU AI Act Articles 9/10/12/13/15/62; HIPAA; SOC 2; CIS AWS Benchmark; 20 IaC policies; SARIF 2.1.0 export; NIST AI RMF 2.0 (73 subcategories + Gen AI profile, NIST AI 600-1); ISO/IEC 42001:2023 (47 controls, certifiable AI management system); DORA (38 controls, EU financial entities, in force Jan 2025); FedRAMP Rev 5 (248 controls across 18 families, Low/Moderate/High baselines); PCI DSS 4.0 (83 sub-requirements, mandatory Mar 2025, customized approach notation); cross-framework traceability matrix with 180+ control links
FinOps AWS CUR ingestion via DuckDB; FOCUS 1.3 (all 33 columns + AI/LLM rows); RI/SP optimization; right-sizing with CloudWatch; carbon emissions; savings executive report
Observability OpenTelemetry gen_ai.* conventions; 8 Prometheus metrics; structlog JSON; Grafana eaa_platform + eaa_cost dashboards; Jaeger traces; OTEL Collector
Audit SHA-256 Merkle chain; reasoning traces as Annex IV evidence; SARIF 2.1.0 to GitHub Security tab; 72-hour Article 62 incident tracking
AI Governance Extended-thinking reasoning trace persistence; Citations API grounded evidence; bias detection; NIST AI RMF scoring; EU AI Act Annex III classification

Cost Optimization — ~95% Savings Story

The core/ optimization layer applies four levers automatically:

Lever Mechanism Saving
Complexity routing ModelRouter scores each task; simple tasks go to Haiku 4.5 ($0.25/MTok input) not Opus 4.7 ($15/MTok) Up to 60× on worker tasks
Result cache SQLite-backed ResultCache returns identical results without a second API call 100% on cache hits
Batch coalescing BatchCoalescer auto-submits accumulated requests to the Anthropic Batch API 50% discount on batched calls
Prompt caching 5-min ephemeral on all system prompts; 1-hour on executive chat ~85–90% on repeat pipelines

Combined baseline: a 1,000-workload 6R scan at all-Opus-4.7 list price costs ~$150. With routing + batching + caching it drops to ~$7–10.


Performance & Cost

Latency benchmarks

Operation p50 p95 Notes
6R classification (Haiku 4.5, cached system prompt) 680 ms 1.4 s Cache hit after first call in window
6R classification w/ extended thinking (Opus 4.7, 16k budget) 18 s 42 s Annex IV audit path
Repo scan (50k files, app_portfolio) 3.8 s 7.2 s Parallel I/O, no AI calls
IaC policy scan (200 resources) 480 ms 1.1 s 20 policies, pure Python
CVE scan (500 deps, OSV batched) 2.1 s 3.9 s Single batched API call
FinOps RI/SP recommendation (10k workloads) 4.2 s 9.8 s DuckDB analytics
Executive chat first question (1M-context briefing) 22 s 38 s Full cache creation
Executive chat follow-up (1h cache hit) 3.1 s 6.4 s 90%+ cost reduction vs first call

Cost benchmarks (per-pipeline, Claude list prices)

Scenario Cost per run vs. baseline
Baseline (all Opus 4.7, no caching, no batch) $0.82 1.00x
+ Prompt caching (5-min ephemeral on system prompts) $0.31 0.38x
+ Model router (Haiku for classification, Sonnet for prose) $0.12 0.15x
+ Batch API on bulk operations $0.07 0.09x
All three combined $0.04 ~95% reduction

Benchmarks are representative estimates based on Anthropic API pricing (April 2026) and typical pipeline sizes. Actual numbers depend on workload characteristics.


See it run in 30 seconds

git clone https://github.com/HunterSpence/enterprise-ai-accelerator
cd enterprise-ai-accelerator
pip install -r requirements.txt
bash examples/run_demos.sh

This runs four end-to-end demos against fixtures in examples/:

  • app_portfolio — scans a Flask sample repo, returns 6R recommendation
  • iac_security — scans Terraform with deliberate violations, returns SARIF
  • sbom — generates CycloneDX SBOM of the sample repo
  • finops_intelligence — analyzes synthetic AWS CUR data, returns savings report

All demos run offline by default (no AWS / Azure / GCP credentials needed). Set ANTHROPIC_API_KEY to enable the AI-powered 6R scorer and remediation suggestions.


This repo runs its own tools

The platform dogfoods itself on every release:

  • CycloneDX SBOMSBOM.cdx.json at the repo root, generated via python -m iac_security sbom .
  • Dependency CVE status — clean, verified via python -m app_portfolio.cve_scanner

EU AI Act Readiness

Enforcement date: August 2, 2026.

The platform is designed to satisfy EU AI Act obligations for high-risk AI system operators:

Article Obligation Platform capability
Article 9 Risk management system Unified 0–100 risk score + per-module traces via risk_aggregator.py
Article 10 Data governance Citations API grounds every compliance claim in cited regulatory source text
Article 12 Record-keeping SHA-256 Merkle chain in ai_audit_trail/ — any tampering detected in O(log n)
Article 13 Transparency Reasoning trace on every extended-thinking call, persisted as Annex IV evidence
Article 15 Accuracy / robustness Extended thinking budget documents model decision process for audit
Article 62 Incident reporting P0–P3 severity ladder + 72-hour deadline tracking in ai_audit_trail/incident_manager.py
Annex IV Technical documentation SARIF 2.1.0 export + structured reasoning trace form a complete Annex IV evidence package

The reasoning-trace + Citations + SARIF combination is not available in any other open-source tool.


How We Compare

Feature Enterprise AI Accelerator AgentLedger AIR Blackbox ai-trace-auditor Langfuse Credo AI
EU AI Act Art.12 Yes (full) Yes Yes (6 articles) Yes (Art.11-13,25) No Yes
SARIF 2.1.0 export Yes No No No No No
OpenTelemetry Yes (native gen_ai.*) No Yes (proxy) Yes (consumer) Yes (v3) No
Tamper-proof chain SHA-256 Merkle SHA-256 SQLite HMAC-SHA256 No No Unknown
Multi-cloud discovery AWS+Azure+GCP+K8s No No No No No
IaC security (20 policies) Yes (CIS/PCI/SOC2/HIPAA) No No No No No
App portfolio scanner Yes (11 languages) No No No No No
Carbon tracking Yes (open coefficients) No No No No No
Python SDK Yes Yes Yes CLI only Yes SaaS
License MIT MIT Apache 2.0 Unknown MIT (core) Proprietary
Cost Free Free Free Free Free (self-host) $50K+/yr

Roadmap

The following are explicitly not yet built. Honest positioning matters.

Gap Status
Multi-tenant RBAC Not built — single-user / single-org only today
React / web dashboard UI Not built — Grafana dashboards for observability only; no app UI
SOC 2 Type II audit Not started — platform itself has not undergone SOC 2 audit
Hyperscaler marketplace listing Not listed on AWS / Azure / GCP Marketplace
Real-time streaming compliance scan In progress — OTEL traces exist; live compliance stream not wired
Multi-region / HA deployment Not documented — single-node only

Repository Structure

enterprise-ai-accelerator/
├── core/                       Anthropic optimization layer
│   ├── ai_client.py            Single Anthropic wrapper with caching + tool-use
│   ├── model_router.py         Complexity-based model selection
│   ├── result_cache.py         SQLite result cache
│   ├── batch_coalescer.py      Auto-coalescing Batch API submitter
│   ├── streaming.py            SSE streaming handler
│   ├── files_api.py            Files API wrapper
│   ├── interleaved_thinking.py Interleaved thinking+tools loop
│   ├── cost_estimator.py       Full cost estimator
│   ├── telemetry.py            OTEL tracer setup
│   ├── prometheus_exporter.py  8 Prometheus metrics
│   └── logging.py              structlog JSON logging
├── cloud_iq/                   AWS infrastructure analysis
│   └── adapters/               Multi-cloud discovery
│       ├── aws.py              boto3 discovery
│       ├── azure.py            azure-mgmt discovery
│       ├── gcp.py              google-cloud discovery
│       ├── kubernetes.py       kubernetes client discovery
│       └── unified.py          UnifiedDiscovery.auto()
├── app_portfolio/              Repository intelligence + 6R scoring
│   ├── cli.py                  CLI entry point
│   ├── analyzer.py             Pipeline coordinator
│   ├── language_detector.py    11-language detector
│   ├── dependency_scanner.py   9 dep manifest formats
│   ├── cve_scanner.py          OSV.dev batch CVE scanner
│   ├── containerization_scorer.py
│   ├── ci_maturity_scorer.py
│   ├── test_coverage_scanner.py
│   └── six_r_scorer.py         Opus 4.7 extended-thinking 6R
├── migration_scout/            6R classification + wave planning
│   ├── assessor.py             AI-native 6R workload classifier
│   ├── dependency_mapper.py    SCC circular dependency resolution
│   ├── wave_planner.py         Monte Carlo wave planner
│   ├── tco_calculator.py       3-year TCO with license elimination
│   ├── batch_classifier.py     Batch API bulk 6R scoring
│   └── thinking_audit.py       Extended-thinking + Annex IV persistence
├── policy_guard/               Multi-framework compliance scanner
│   ├── scanner.py              EU AI Act + HIPAA + SOC2 + PCI-DSS
│   ├── bias_detector.py        Statistical disparate impact analysis
│   ├── sarif_exporter.py       SARIF 2.1.0 → GitHub Security tab
│   ├── incident_response.py    P0–P3 + SLA tracking
│   └── thinking_audit.py       Extended-thinking audit path
├── iac_security/               IaC security + SBOM + drift
│   ├── terraform_parser.py     Terraform HCL parser
│   ├── pulumi_parser.py        Pulumi parser
│   ├── policies.py             20 built-in policies
│   ├── sbom_generator.py       CycloneDX SBOM generator
│   ├── osv_scanner.py          OSV.dev batched CVE scanner
│   ├── drift_detector.py       IaC vs. cloud state diff
│   └── sarif_exporter.py       SARIF 2.1.0 exporter
├── finops_intelligence/        Cloud cost intelligence
│   ├── cur_ingestor.py         AWS CUR ingestion via DuckDB
│   ├── ri_sp_optimizer.py      RI/SP optimizer (80% coverage cap)
│   ├── right_sizer.py          CloudWatch + instance catalog right-sizer
│   ├── carbon_tracker.py       Carbon emissions (open coefficients)
│   └── savings_reporter.py     Executive savings report
├── ai_audit_trail/             EU AI Act logging + NIST AI RMF
│   ├── chain.py                SHA-256 Merkle hash chain
│   ├── eu_ai_act.py            Article 12/62 compliance engine
│   ├── nist_rmf.py             GOVERN/MAP/MEASURE/MANAGE scoring
│   ├── incident_manager.py     P0–P3 + Article 62 deadline tracking
│   ├── decorators.py           Drop-in SDK integrations (5 frameworks)
│   └── sarif_exporter.py       SARIF 2.1.0 export
├── executive_chat/             1M-context CTO Q&A
├── compliance_citations/       Citations API grounded compliance evidence
├── integrations/               Notification + ticketing adapters
│   ├── dispatcher.py           FindingRouter + WebhookDispatcher
│   ├── slack.py / jira.py / servicenow.py / github_app.py
│   ├── teams.py / pagerduty.py / smtp_email.py / github_issue.py
├── observability/              OTEL + Prometheus + Grafana
│   ├── grafana_dashboards/     eaa_platform + eaa_cost dashboards
│   ├── otel-collector.yaml     OTEL Collector config
│   └── docker-compose.obs.yaml One-command observability stack
├── agent_ops/                  Multi-agent orchestrator
├── risk_aggregator.py          Cross-module 0–100 risk score
└── mcp_server.py               18 MCP tools + 4 resources + 4 prompts (stdio + SSE)
└── mcp_transports.py           Transport helpers: run_stdio(), run_sse(), SSE health endpoint

Demo Commands

# App portfolio scan (simplest entry point)
python -m app_portfolio.cli .

# AI governance + EU AI Act (3 enterprise scenarios, no credentials)
python -m ai_audit_trail.demo

# $340K/month cloud spend optimization ($89.4K/month identified)
python -m finops_intelligence.demo

# 75-workload migration plan, Oracle $420K/yr license elimination
python -m migration_scout.demo

# EU AI Act compliance scanner (Fortune 500 hiring AI + healthcare AI)
python -m policy_guard.demo

# AWS infrastructure analysis ($47,200/month waste identified)
python -m cloud_iq.demo

# Bring up full observability stack (Prometheus + Grafana + Jaeger)
cd observability && docker compose -f docker-compose.obs.yaml up -d

# MCP server — stdio (Claude Code / Claude Desktop local)
python mcp_server.py

# MCP server — SSE transport (remote / CI agent access)
python mcp_server.py --transport sse --host 0.0.0.0 --port 8765

# Health check (SSE mode only)
curl http://localhost:8765/health

See docs/DEMO.md for the 5-minute exec demo, 15-minute technical walkthrough, and 3-minute interview pitch.


Why This Matters Now

EU AI Act — August 2, 2026: High-risk AI system obligations (Articles 8–25) become enforceable. Logging, documentation, human oversight, and incident reporting requirements apply. Article 62 requires serious incident reporting within 72 hours. Non-compliance: up to 3% of global annual turnover.

AWS Migration Hub closure — November 7, 2025: The standard OSS migration planning tool is gone. AWS Transform covers only .NET and mainframe. The market gap for general-purpose migration intelligence is open.

FOCUS 1.3 adoption: Now the basis for multi-cloud billing normalization across enterprise FinOps platforms. Organizations without FOCUS-compliant tooling face manual data transformation across every cloud billing export.


Requirements

Python 3.11+
anthropic>=0.69.0

Full dependency list: requirements.txt. Key additions in v0.2.0: boto3, azure-mgmt-compute, azure-mgmt-resource, google-cloud-compute, kubernetes, opentelemetry-sdk, opentelemetry-exporter-otlp, prometheus-client, python-hcl2, cyclonedx-python-lib, packageurl-python, PyJWT, cryptography, slack-sdk, jira.

All dependencies are OSS (Apache 2.0 / MIT). Zero paid SaaS services.


MCP 2.0 Surface

The MCP server exposes three capability tiers:

Tools (18 — callable by the client)

Tool Module
audit_log_decision, get_compliance_status, export_sarif, get_audit_chain AIAuditTrail
cloudiq_analyze_environment CloudIQ
migration_assess_workload, migration_bulk_classify, migration_generate_wave_plan MigrationScout
finops_explain_anomaly, finops_bulk_explain FinOps
policyguard_scan_iac, policyguard_audit_policy, policyguard_audit_bias PolicyGuard
executive_ask ExecutiveChat
compliance_cite_question ComplianceCitations
list_models, platform_capabilities, risk_aggregate_score Platform

Resources (4 — streamable, not inlined)

URI Content
audit-trail://recent Last 50 audit decisions as JSON
audit-trail://chain-verify Merkle chain verification result
scan-results://{scan_id} Full IaC scan result by ID (populated by policyguard_scan_iac)
compliance://frameworks Supported regulatory frameworks (7 total)
policy-catalog://iac 20-policy IaC catalog (CIS AWS, SOC 2, GDPR, PCI-DSS)

Prompts (4 — reusable templates)

Prompt Args Purpose
audit-terraform path, environment Scan + narrate IaC findings
classify-workload-6r workload_json 6R classification with reasoning trace
assess-bias dataset_summary EU AI Act Article 10 bias audit
executive-briefing scan_results_json Board-level CTO briefing

Transports

# stdio (default — Claude Code, Claude Desktop local)
python mcp_server.py

# SSE (network-accessible — remote Claude Desktop, CI pipelines)
python mcp_server.py --transport sse --host 0.0.0.0 --port 8765

Claude Desktop SSE config:

{
  "mcpServers": {
    "enterprise-ai-accelerator": {
      "url": "http://localhost:8765/sse",
      "transport": "sse"
    }
  }
}

SSE mode exposes GET /health:

{"status": "ok", "tools": 18, "resources": 5, "prompts": 4, "uptime_s": 42.1}

Author

Hunter Spence 4 years at Accenture, Infrastructure Transformation (CL-9). Delivered cloud migration engagements across enterprise clients. AWS Certified Cloud Practitioner.

LinkedIn · Email · VantaWeb


Contributing

Pull requests welcome. See CONTRIBUTING.md for the contribution guide and code style.


License

MIT. Use it, extend it, white-label it. See LICENSE.

Built because the gap between what Big 4 firms charge and what the technology can do autonomously is no longer defensible.

About

AI-native unified cloud governance platform — multi-cloud discovery, 6R migration planning, IaC security, FinOps, compliance audit, and executive AI chat. Built on Claude Opus 4.7. EU AI Act Annex IV ready. Zero paid SaaS dependencies.

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages