Tim Wolfe
AI Platform Architect · A2A · Agents · MCP · Secure Systems
Advisory
Private A2A · Agent · MCP Advisory Practice

Governance. Compliance. Security. Architecture.

A2A · Agents · MCP

Production-grade AI agent systems for regulated enterprises — where "it mostly works" isn't an answer, and a regulator can audit any decision down to the hash.

The thesis. Most enterprise AI deployments fail the audit they haven't had yet. I build the stack that passes it — compile-time governance, runtime enforcement, post-quantum A2A transport, and compliance automation across 22 regulatory frameworks. Agents are governed before they run, validated before they deploy, and recallable in under one turn.
Bastion · Deployment Envelopes
Four SKUs · one platform
Citadel
Hosted SaaS
Fortune 500 · GovCloud
Stronghold
Enterprise On-Prem
Regulated · Full stack
Enclave
DoD Air-Gap
IL4/IL5 · SCIF · ATO
Sovereign
Classified IL6
IC · CLI · Zero network

Six practices, one operator. Click any to expand. Filter by sector to narrow the list.

Agent Governance & Constitutional AI
Production AI agents with hard governance guarantees — compile-time enforcement, runtime validation, instant recall.
Flagship Castellan OWASP Agentic
+
What ships
Constitutional policy compilation · 6-stage deterministic pipeline · Rust runtime enforcement · 30-check policy gate · kill switch propagation.
Frameworks supported
LangGraph · CrewAI · AutoGen · Haystack — governance embedded inline via --governed export.
Audit posture
Every decision traceable to the specific policy rule and spec field that triggered it. Zero LLM in the enforcement path.
Scope this engagement →
Compliance & Regulatory Navigation
FedRAMP, NIST, HIPAA, ISO 42001, EU AI Act — mapped to your AI systems with automated SSP generation.
22 Frameworks OSCAL JSON SPRS scoring
+
Federal
FedRAMP · NIST 800-53/171/207 · CMMC · DoD AI RMF · DISA STIG · FDA PCCP.
Industry
HIPAA · PCI-DSS · SOC 2 · ISO 42001 · ISO 27001 · CSA Agent Trust Framework.
International
EU AI Act · GDPR · FZ-152 · CNSA 2.0 · CNSSP-12 alignment.
Scope this engagement →
MCP Security & Supply Chain
Harden agent tool chains against poisoning, rug pulls, and exploit chaining. 20+ detection patterns.
OpenClaw SHA-384 Level 6 hardened
+
Design-time defense
Tool description poisoning scanner · 20 regex patterns · 4 severity levels · shadowing detection.
Runtime protection
Tool integrity re-verification · 4-dimensional trust scoring · PII redaction at call-site · per-channel policy.
Arsenal coverage
52+ hardened Rust MCP servers across 9 DoD functional areas · SCIF-grade isolation.
Scope this engagement →
DoD Hardening & TS/SCI Deployment
AI agents compiled to DoD specification — SCIF, air-gap, classified network-ready with full ATO packages.
Defense Only TS/SCI CNSA 2.0
+
ATO deliverables
SSP (OSCAL JSON, NIST 800-53 Rev 5) · PPS · CMR · STIG checklists (V-ID, CAT I/II/III) · CycloneDX AI SBOM.
Crypto & identity
FIPS 140-3 · post-quantum (ML-DSA-87, SLH-DSA-256) · SPIFFE/SPIRE · CAC authentication · 60-second token TTL.
Deployment posture
Air-gapped · classification banners UNCLASSIFIED → TS//SCI · zero cloud dependencies · Level 6 hardened Rust.
Scope this engagement →
AI SDLC Automation & Gates
Deterministic pipelines producing SHIP / REVIEW / BLOCK verdicts at the merge gate.
Charlotte 54 checks CI/CD native
+
Pipeline stages
Validate → Compile → Scan → Lint → Test. Context engineering for system prompts, Claude Code skills, MCP servers.
Security coverage
OWASP Top 10 Agentic · prompt injection filtering · PII leakage · MCP tool poisoning · 21 categories.
CI integration
castellan ci-diff posts APPROVE / REVIEW / BLOCK as PR comment. GitHub · GitLab · Jenkins.
Scope this engagement →
Enterprise AI Platform Architecture
Multi-provider orchestration with governance woven into the design — ReAct runtime, A2A protocol, circuit breakers.
3 providers Post-quantum A2A Circuit breaker
+
Runtime design
Async-first ReAct (Observe → Think → Act → Validate) · Anthropic · OpenAI · Google · MCP client · RAG · LRU+TTL caching.
Multi-agent patterns
Supervisor · pipeline · broadcast · peer collaboration · delegation depth limiting · dynamic routing.
Cost governance
Per-run budget enforcement · model routing by cost/capability · circuit breaker states · per-tenant tracking.
Scope this engagement →

The regulatory fluency that underpins every engagement. Grouped by jurisdiction.

Why this matters. Most AI deployments are built first and compliance-fit later — a costly inversion. These frameworks are baked into my toolchain: live codebase scanning, automated SSP generation, and behavioral policy templates pre-aligned to each.
Federal · U.S. Government
7 frameworks
FedRAMP Low / Moderate / High baselines · OSCAL JSON SSP
NIST 800-53 Rev 5 · full control family mapping
NIST 800-171 110 CUI requirements · SPRS scoring
NIST 800-207 Zero Trust Architecture
CMMC Levels 1–3 · DoD supply chain
NIST AI RMF AI Risk Management Framework
FDA PCCP Predetermined Change Control · 7 categories
Defense · Classified & Tactical
6 frameworks
DoD AI RMF Department of Defense AI Risk Framework
DISA STIG V-ID mapping · CAT I/II/III · CKL export
DoD IL4 / IL5 Impact Level classifications
FIPS 140-3 Cryptographic module validation
CNSA 2.0 Commercial National Security Algorithm
CNSSP-12 TS/SCI classified systems floor
Industry · Commercial Regulated
6 frameworks
HIPAA Healthcare · PHI handling & audit
PCI-DSS Payment Card Industry Data Security
SOC 2 Type II · Trust Services Criteria
ISO 42001 AI Management System · 32 clauses
ISO 27001 Information Security Management
OWASP Agentic Top 10 for Agentic Applications
International · Cross-Border
3 frameworks
EU AI Act European Union AI regulation
GDPR General Data Protection Regulation
FZ-152 Russia · personal data localization

Drawbridge Plus — post-quantum ready A2A transport. Triple-signature hybrid, CNSA 2.0 on the wire, downgrade-strict verification.

The crypto foundation. Post-quantum cryptography isn't a roadmap item — it's already shipping. Every agent-to-agent message signs with three independent layers (classical + PQ lattice + PQ hash-based), encrypts with CNSA 2.0 mandated AEAD, and authenticates with downgrade-strict verification. Live TCP round-trip proven end-to-end.
Security properties
Triple-Signature Hybrid Signing
Every envelope signs with three independent signature schemes. Classical + PQ lattice + PQ hash-based. A break in any single algorithm does not forge.
Ed25519 · ML-DSA-87 · SLH-DSA-256
Downgrade-Strict Verify Enforcement
If a bundle declares a PQ layer, the signature MUST carry it. Stripping layers is a hard verification failure, not a silent fallback.
has_pqc · has_slh_dsa bundle gates
CNSA 2.0 Wire Crypto Compliance
NSA Commercial National Security Algorithm suite on every wire payload. The mandated baseline for U.S. national security systems.
AES-256-GCM · SHA-384 · HMAC-SHA384 · HKDF-SHA384
Forward Secrecy + Replay Session
Fresh ephemeral hybrid-KEM keypair per call. Timestamp + nonce replay window. Per-operator rate limit keyed on fingerprint, not IP.
X25519 + ML-KEM-1024 · ±5 min freshness
Algorithm inventory
Layer Primitive Standard
Sig Ed25519 FIPS 186-5 · RFC 8032
Sig ML-DSA-87 FIPS 204 (lattice)
Sig SLH-DSA-256 FIPS 205 (hash-based)
KEM X25519 RFC 7748 (classical DH)
KEM ML-KEM-1024 FIPS 203 (HKDF combined)
AEAD AES-256-GCM FIPS 197 · SP 800-38D
Hash SHA-384 FIPS 180-4
MAC HMAC-SHA384 FIPS 198-1 · RFC 2104
KDF HKDF-SHA384 RFC 5869
At-rest Argon2id OWASP 2024 (identity KEK)
At-rest XChaCha20-Poly1305 RFC 8439 ext (192-bit nonce)
Standards alignment
CNSA 2.0 Wire Crypto
FIPS 140-3 Module Validation
FIPS 203/204/205 Post-Quantum

Three ways to work together, scoped to the shape of the problem. Most clients start with Assessment.

Assessment
2–4 weeks · fixed fee
Diagnose where you stand before committing to a build. Architecture review, governance-gap analysis, compliance mapping, remediation roadmap.
  • Architecture & compliance review
  • Governance-gap diagnostic
  • Framework mapping
  • Remediation roadmap
  • Executive briefing
Build
6–20 weeks · milestone-based
Implementation engagement. Stand up the governance stack, integrate compliance automation, ship with your team trained to operate it.
  • Governance stack
  • Compliance automation
  • MCP hardening
  • DoD packaging
  • CI/CD gates
  • Team training
Retainer
Ongoing · monthly
Technical advisory board seat. Monthly strategic review, continuous compliance monitoring, on-call incident response.
  • Advisory board seat
  • Drift monitoring
  • Framework tracking
  • Incident response
  • Quarterly review
Los Altos, California
Tim Wolfe
AI Platform Architect · A2A · Agents · MCP · Secure Systems

Fastest path is email with a one-paragraph description of the problem. I'll respond within 48 hours with a scoped proposal or a referral.

Email rtwolfe@gmail.com Direct 650.390.5003 LinkedIn /in/timwolfe Telegram @timwolfe GitHub /rtwolfe/portfolio
⚬ Available for select engagements
Tim Wolfe Advisory DoD · NIST-aligned · CSPO · A-CSM