Hi, I'm Pabasara.
Senior Software Engineer at WSO2, specialising in Identity and Access Management and enterprise integration. I write about IAM, Agentic AI, frontend engineering, software architecture, and what breaks in production.
What people say
All recommendations →-
Pabasara demonstrated strong expertise in API Gateway, integration, and middleware technologies, and played a key role in delivering scalable and reliable integration solutions in a complex banking environment.
Oan Ali Enterprise & Digital Transformation Leader, Chief Architect Read in full → -
He delivered at a pace our testers couldn’t keep up with and he was always super easy to deal with. I have since cited him in numerous interviews and conversations with peers and his code changes still provide a template for our in house developers to copy.
Martin Jones Integration Architect, University of Edinburgh Read in full →
GitHub activity
GitHub profile ↗Recent writing
Full archive →-
Turning WSO2 Identity Server's SCIM 2.0 APIs into a governed MCP server with the WSO2 AI Gateway
How to expose the SCIM 2.0 APIs of WSO2 Identity Server 7.3 as an MCP server, secure it with Identity Server's own tokens at the WSO2 AI Gateway, and let Claude Desktop manage users and groups.
-
mcp-remote Crashed When the OAuth Port Was Taken, Here's How I Fixed It
An open source contribution to mcp-remote: recovering from EADDRINUSE on the OAuth callback port without silently breaking the registered redirect_uri.
-
Business Rules Don't Have to Live in a Wiki Page: Enforcing Them with a Roslyn Analyzer
A business rule that lived only in developers' memory, pushed down into the compiler with a .NET Roslyn analyzer that warns on every rule violation
-
How the Upcoming MCP Auth Spec Tightens Up (PKCE Won't Save You)
The 2026-07-28 MCP authorization release candidate brings six hardening and interoperability fixes: issuer validation, per-issuer credentials, refresh token guidance, and more. Explained in plain terms.
-
Sorting 70,000 Photos Without Losing the Weekend
How I used OpenAI's CLIP model to classify 70,000 photos into 13 categories locally in three hours — no cloud, no API, no labelled data.