Pick another character
TECH LEAD ROUTE

The agentic stack, and what I've built at each layer

Every layer of a modern agent system, matched to the work that proves it. Tap an app to light up the layers it covers.

People and channels
Web, chat and voice, wherever people already work.
Mini StepsMini CallsMini OpsLowes
Partner agents
A client’s own agents on other platforms.
Next
Orchestration and agents
Plans the work, hands steps to specialist agents and pauses for a person to approve.
Ship KeysMini CallsMini Engine
Agent-to-agent (A2A)
Outside agents hand work to mine, and back.
Mini OpsShip KeysNext
LLMs
AI models matched to how complex each query is.
Mini Navigator
Memory and context
Remembers goals across conversations and finds the right documents for each task.
Mini StepsMini Engine
Tools via MCP
My own MCP servers and clients.
Ship KeysMini StepsMini CallsMini Navigator
Business systems and data
Connects to the systems a business already runs, such as Intelliflo Office, the FCA Register and identity checks.
Mini OpsLowes
Guardrails
A person approves before anything is saved or sent.
Mini Steps
Evals and quality
Launch checks, quality gates and evals.
Ship KeysMini Engine
Cost control
Spending caps, so AI costs stay predictable.
Mini Steps
Security
Permission checks, audit logs and strong privacy controls.
Mini Navigator
HOW A MODERN AGENT SYSTEM FITS TOGETHER
ACROSS EVERY LAYER
THE WHOLE STACKEvery layer backed by real work

Each box is a layer of a modern agent system. The tags show which of my apps proves it. Tap an app above to light up its layers.

KEYBuilt or in progressNext on the roadmapLayers the chosen app covers

The apps, one by one

Ship Keys

Agents in customers’ CI pipelines

Launch checks for AI-built apps. Autonomous agents run in customers’ CI pipelines, and its own MCP server and Claude connector bring it into the tools people already use. Private beta.

Mini Steps

Goals that AI remembers, with the owner in charge

Persistent goals and context across conversations, owner-approved actions and per-call metering with spend limits. Two MCP servers, including an OAuth 2.1 Claude connector for end-to-end voice interaction. Private beta.

Mini Calls

Voice agents with secured actions

End-to-end voice control for your business, with AI that keeps track of every conversation. Agents hand calls to each other, use tools through MCP, and each action is secured before it runs. Built, on Cloud Run.

Mini Navigator

Privacy-first design for sensitive data

Records for families of children with special educational needs and disabilities (SEND), with strong privacy controls, the right AI model for each query, and its own MCP server. In development.

Mini Engine

RAG and LangGraph, with people in the loop

Drafts review packs through RAG and LangGraph pipelines, with human approval checkpoints and evals. A prototype, and my current work.

Mini Ops

Scheduled AI reports for small businesses

AI reports on a schedule, linked to the tools a business already uses, with actions that wait for the owner to approve them. A prototype, and the natural fit for A2A.

Lowes (FCA-regulated)

Regulated enterprise at scale

My 11 years’ experience in regulated financial services includes seven years running technology at Lowes Financial Management, an FCA-regulated financial advisory firm with over £1bn of assets under management: 44 applications and 45 databases, serving 80+ staff and 5,000+ clients. Accountable for six-monthly client valuation reports and annual costs and charges reporting for over 4,000 clients, ensuring accuracy before issue. Also served as Deputy Data Protection Officer. Its onboarding platform, rolled out to over 80 staff (including 18 advisers), cut application processing time by over 90%.

Work with me

Email me, or book a meeting straight into my calendar.

Book a meeting

A free 30-minute call. Pick a time that suits you.

Book a meeting (opens in a new tab)