[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fYGE9ZqM9J-bLNHUdAURpR3_s7JWr3qu1LKAlnEdDwCE":3},{"id":4,"slug":5,"title":6,"projectType":7,"client":8,"tools":9,"year":10,"summary":11,"problem":12,"approach":13,"outcome":14,"previewUrl":15,"repoUrl":15,"thumbnailUrl":16,"images":17,"sortOrder":18,"active":19,"next":20},5,"azure-devops-mcp-server","Azure DevOps MCP Server","Developer Tooling","TransPerfect — internal developer platform","TypeScript, Node.js, Model Context Protocol, Azure DevOps REST API, PowerShell","2025","An MCP server that exposes Azure DevOps as structured tools, so an LLM can carry out routine developer workflows through tool calling instead of a human clicking through the portal.","A lot of day-to-day engineering time went into Azure DevOps chores — checking build and release state, moving work items, pulling pull request context, setting up environments. Each one is small, but they interrupt whatever you were actually doing, and the Azure DevOps UI makes you navigate several screens for answers that are one API call away.","I built a Model Context Protocol server in TypeScript and Node.js that wraps the Azure DevOps REST API. Rather than exposing raw endpoints, each operation is published as a tool with an explicit schema — a clear name, typed arguments, and a description an LLM can reason about — so the model picks the right call and gets a validated payload back. Auth and scoping happen server-side, so the model never handles tokens, and destructive operations are deliberately kept out of the tool surface. PowerShell scripting handles the Windows-side environment setup steps that sit outside the REST API.","Now a daily-use internal tool — routine Azure DevOps work runs through the assistant rather than the portal, and adding a new automated workflow means defining one tool instead of writing a bespoke script. It also became the reference implementation for how we expose internal systems to agentic tooling.","","\u002Fimg\u002Fprojects\u002Fcover-mcp-server.svg","[]",1,true,{"slug":21,"title":22},"rag-knowledge-assistant","AI-Powered Knowledge Assistant"]