Implementation · MCP
A payments infrastructure company
Exposed a legacy payments portal to AI agents through an MCP server — without a risky big-bang rewrite.
The pipeline in action
- The situation
- A payments company running a legacy customer portal on an aging LAMP stack needed to expose its capabilities to AI agents without a risky big-bang rewrite.
- What we did
- Ran a discovery process across the existing portal and REST API surface, then designed and specified an MCP server in TypeScript with mock, sandbox, and live modes and a full audit trail. Recommended a parallel path — ship the AI interface now, migrate the portal incrementally using a strangler-fig pattern — rather than sequencing one behind the other.
- Engagement shape
- Discovery → architecture → phased build
More case studies
Agentic Development · Enablement
A global cybersecurity company
A multi-cohort AI enablement program that moved 60–80 engineering managers from experimenting with AI coding tools to using them by default.
Training · Enablement
A behavioural biometrics company
A two-day GenAI intensive that turned program managers from AI observers into daily practitioners.