Bryan Thompson

Open-Source MCP Tooling Developer

I build tooling that makes MCP servers testable before they ship: assessment harnesses, protocol-conformance suites, and a development harness that keeps agents honest — gates that run before an agent emits text a human will act on, and a fleet whose members are built not to take each other's word. All of it in public repositories. Before this: a career in enterprise systems — SQL Server administration, then Fortune 500 operations at PayPal and Fiserv, where the job was keeping up systems other people had already bet on. That is the same job, with the failure modes moved.

From enterprise systems to MCP infrastructure

Enterprise reliability → production AI → open-source MCP infrastructure. Starting in SQL Server administration, I built up through Fortune 500 operations — high-volume transaction systems at BadgePass, leading a 15-engineer team across 200+ sites at Elior, becoming DNA CCM subject-matter expert at Fiserv, and driving SAP Bank Analyzer integration at PayPal. That foundation is what I now apply to MCP infrastructure and quality tooling.

Current: MCP tooling

  • • Open-source MCP tooling
  • • Assessment and conformance harnesses
  • • Agent verification harness
  • • Vector search infrastructure

AI transition

  • • Triepod.ai: personal AI tooling and research projects
  • • Lodestar: financial M&A systems
  • • First production MCP servers
  • • Vector database architecture

Fortune 500

  • • PayPal: SAP Bank Analyzer
  • • Fiserv: DNA CCM SME, 50+ engineers trained
  • • SOC2/PCI compliance at scale
  • • Incident response and change control

Enterprise foundation

  • • Elior: led 15 engineers, 200+ sites
  • • BadgePass: Fortune 500, 50K+ daily
  • • Healthcare systems under HIPAA
  • • VMware, SQL Server, automation

My Developer Journey

From enterprise business analysis to AI infrastructure. Each step is here because it changed how I build, including the one that was not a career decision.

Foundation

IT Foundation Years

Started as IT support, progressed to Lead IT role. Built foundational understanding of enterprise systems, user needs, and technical problem-solving.

Transition

Software Support to Analysis

Moved from software support into merger and process analysis work. Built the habits that carried forward: process analysis, data integration, and business transformation.

Enterprise

PayPal & Fiserv

Building automation for manual back-office processes on Fortune 500 payment systems. Enterprise patterns and financial system integration.

The turn

Life-Changing Event

COVID took my leg but gave me perspective. Used recovery time to dive deep into AI and machine learning, turning challenge into opportunity.

Deep dive

AI Deep Dive

Intensive self-study in machine learning, LLMs, and AI architectures. Built first automation tools and discovered passion for accessible AI development.

MCP, early

Early MCP development

Early adopter of Model Context Protocol. Developed expertise by building and running production MCP servers, and contributing to the ecosystem.

Directory platform

ToolNexus Platform Launch

Created an MCP directory platform and the cataloguing around it, with a focus on GPU-accelerated inference.

MCP Tooling

Open-Source MCP Tooling Developer

Building open-source automation infrastructure for the MCP ecosystem: assessment harnesses, a verification harness for agent work, native module debugging, OAuth patterns and bundle optimisation. Applying a career of enterprise reliability practice to AI infrastructure.

Now

Tooling, and the harness around it

Ongoing open-source work: assessment and conformance tooling for MCP servers, the gates that check agent work before a human acts on it, and the trust boundary between agents in a multi-agent fleet.

What the work actually is

I build tooling that makes automated systems trustworthy — and I do the deep technical work that the tooling exists to make repeatable.

Keeping the agents honest

Gates that run before an agent emits text a human will act on, measurement of what an agent actually does rather than what it reports doing, and a fleet built so that no member takes another's word for anything.

Platform & security

  • CVE-2025-53110 response — a path-prefix escape in a widely deployed reference server
  • • Native module compatibility debugging (Prisma, SQLite, AWS)
  • • OAuth pattern validation and implementation guidance
  • • Runtime footprint reduction for Python server bundles

Developer enablement

  • • Documentation-first technical guidance
  • • Failure pattern libraries and best-practice documentation
  • • Patterns that were adopted as ecosystem standards
  • • Neutral, observation-based technical communication

The turn into AI

The turn was not a career plan. When COVID affected my mobility, I channelled recovery time into learning AI and machine learning in earnest. It was not simply picking up new tools — it was recognising that AI could amplify the same system-optimisation work I had been doing across a career in enterprise systems, and that the reliability instincts came with me.

Where that went

  • • Model Context Protocol server development
  • • Vector database optimisation (Qdrant)
  • • Multi-agent orchestration, and the trust model between agents
  • • Runtime and bundle performance engineering
  • • Real-time transport and streaming interfaces

Certifications & Education

AI & MCP
  • Building with the Claude API
    Anthropic
  • Introduction to Model Context Protocol
    Anthropic
  • Model Context Protocol: Advanced Topics
    Anthropic
Professional Development
  • AI Fluency for Students
    Wharton Online
  • AI Fluency Framework & Foundations
    Wharton Online
Education
Computer Engineering
Technical Foundation

Continuous Learning: These certifications represent focused study in AI systems, particularly the Model Context Protocol ecosystem, building on a foundation of Computer Engineering and a career in enterprise systems.

Where to go next

Follow the work

How the directory pipelines are built, and the public record behind every number on this site.