fde-mastery MCP server

FDE Mastery

FDE Mastery is a governed execution platform for teams building, integrating, and operating enterprise AI agents and workflows.

It provides reusable identity and tenancy boundaries, agent execution, durable workflows, policy enforcement, human approvals, model and tool gateways, evaluations, observability, and operational controls across eight enterprise domains.

Architecture at a glance

The diagram describes platform boundaries; it is not a screenshot or live deployment demonstration. See the architecture guide.

Why FDE Mastery

Concern	Platform approach
Identity and tenancy	Tenant-aware identity context, scoped authorization and PostgreSQL RLS foundations
Agent execution	Versioned agent lifecycle, budgets, checkpoints and durable workflow boundaries
Consequential actions	Independent trust/policy controls, risk tiers and expiring human approvals
Tool and model access	Registries, capability boundaries, policy checks and provider adapters
Reliability	Leased work, retries, recovery, idempotency and incident lifecycle contracts
Evaluation	Golden and adversarial datasets, quality/safety gates and drift signals
Operations	OpenTelemetry-compatible observability, tenant-aware FinOps and deployment controls
Supply chain	CI security checks, SBOM generation, image signing and build provenance

FDE Mastery is a modular monolith with explicit control-plane, data-plane and trust-plane boundaries. It does not split components into microservices before independent scaling or isolation is justified.

Quick start

Run the repository tests with local in-memory and mock-provider settings. No production credentials are required.

git clone https://github.com/LloydCoder/fde-mastery.git && cd fde-mastery/packages/platform-core
python -m venv .venv && source .venv/bin/activate
python -m pip install -e ".[test]"
FDE_STORAGE_BACKEND=memory FDE_RATE_LIMIT_BACKEND=memory FDE_MONTH1_PROVIDER=mock MOCK_LLM=true FDE_ENVIRONMENT=test python -m pytest -q

Windows PowerShell activation uses .venv\Scripts\Activate.ps1. Python 3.11 matches the primary CI runtime; package metadata supports Python 3.10 and newer.

Installation

Tool	Requirement
Python	3.10+; Python 3.11 is the CI baseline
Git	Required to clone
Docker Compose	Optional, for container deployment
Node.js	20+ for the TypeScript SDK

From packages/platform-core, install the platform and quality tools:

python -m pip install -e ".[test,quality,security,observability,sbom]"

Optional dependency groups are defined in pyproject.toml. The production-oriented Compose definition is deployment/docker-compose.production.yml. It requires deployment-specific PostgreSQL, OIDC, application/admin key and managed-secret configuration. Follow the deployment guide; never use example credentials in production.

Usage

Inspect platform capabilities

From the repository root:

PYTHONPATH=packages/platform-core python -m platformctl manifest
PYTHONPATH=packages/platform-core python -m platformctl doctor

Run the API locally

After installing the package, from packages/platform-core:

FDE_STORAGE_BACKEND=memory FDE_RATE_LIMIT_BACKEND=memory FDE_MONTH1_PROVIDER=mock MOCK_LLM=true FDE_ENVIRONMENT=test python -m uvicorn deployment.api_gateway.main:app --host 127.0.0.1 --port 8000

In another terminal:

curl --fail http://127.0.0.1:8000/v1/health

See API reference for authentication, tenant binding, idempotency and request metadata. Do not send customer data to a local test configuration.

SDKs

SDK	Runtime	Source
Python	Python 3.10+; standard-library HTTP client	sdks/python
TypeScript	Node.js 20+	sdks/typescript

Configuration

The tracked API gateway example is .env.example. It contains placeholders only. Copy values into an untracked environment file and replace every placeholder before use.

Variable	Template example	Purpose
FDE_API_KEYS	Placeholder	Application API keys
FDE_ADMIN_API_KEYS	Placeholder	Separate administrator API keys
FDE_STORAGE_BACKEND	memory	Storage adapter
FDE_DATABASE_URL	Local PostgreSQL URL	Durable database connection
FDE_DB_CREATE_TABLES	false	Automatic table creation
FDE_RATE_LIMIT_BACKEND	memory	Rate-limit backend
FDE_REDIS_URL	Local Redis URL	Redis connection
FDE_RATE_LIMIT_WINDOW_SECONDS	60	Rate-limit window
FDE_RATE_LIMIT_REQUESTS	60	Request allowance
FDE_MAX_BODY_BYTES	262144	Maximum request body size

See Configuration reference. Template values are not production recommendations.

Platform domains and capabilities

Domain	Example workflow
Cybersecurity	SIEM alert enrichment and SOC triage
Finance	Transaction risk and governance
HealthTech	Compliance and triage support
Logistics	Supply-chain risk analysis
Legal	Contract risk workflows
RevOps	Pipeline and account operations
Procurement	Supplier risk and quote comparison
Custom	Tenant-defined workflows and policies

The platform provides versioned agent/tool/model/policy registries, durable workflows, tenant-scoped events, evaluation, incident management, AI FinOps, integration and interoperability boundaries, and deployment profiles.

High-impact actions remain subject to policy and human approval. Procurement is recommendation-oriented: the platform does not autonomously award suppliers, create purchase orders, modify vendor master data or approve spend.

These are reusable platform capabilities and reference integrations. They are not a security certification, regulatory approval or evidence of a live customer deployment.

Documentation

Guide	Description
Documentation index	Documentation map and reading paths
Getting started	Local setup and first checks
Architecture	Platform boundaries and dependency rules
Architecture decisions	Accepted architecture decision records
API and SDKs	Versioned API and client contracts
Configuration	Runtime environment settings
Deployment	Deployment model and operations
Production readiness	Eight-gate promotion model
Build status	Historical implementation and verification ledger
Security standards	Engineering control references
AI threat model	AI-specific threat considerations

Contributing

Issues and pull requests are welcome. Read CONTRIBUTING.md, follow the Code of Conduct, and use the provided templates.

License and acknowledgements

FDE Mastery is distributed under the Apache License 2.0. The project uses and is informed by OWASP, NIST AI RMF, OpenTelemetry, CycloneDX, Sigstore Cosign and SLSA provenance.

Responsible use

Domain examples use synthetic or illustrative data. They are not financial, medical or legal advice. Customer deployments require their own threat model, data-protection decisions, integration validation, evaluation evidence, security review, operational runbooks and approval policy.

• Defects: GitHub Issues.
• Vulnerabilities: follow SECURITY.md; do not publish exploit details.
• Setup: see Getting started and Configuration.
• Deployment: see DEPLOYMENT.md.