About us
A community for software engineers exploring how AI can supercharge the way we build and operate software. We focus on practical, production-ready uses of AI in engineering workflows, including:
- AI-assisted coding – Claude Code, GitHub Copilot, and beyond
- AI-driven dev environments & orchestration – MCP servers, CrewAI, LangGraph, PydanticAI
- Automated testing & CI/CD – accelerating quality and delivery
- AI-powered operations & monitoring – intelligent alerts, diagnostics, and remediation
- Infrastructure automation – provisioning, scaling, and management with AI
This is not a data science meetup — we focus on the engineering side of AI: tools, automation, and systems that make software development faster, smarter, and more reliable.
Join us in Manchester, the #1 AI-ready city in the UK, and be part of the future of AI in software engineering!
Upcoming events
3
David White
Matillion, Two, New Bailey St, Stanley St, Salford M3 5GS, Manchester, GBThis month Practically AI brings together two very different but highly practical perspectives on building AI systems that work in the real world.
We'll start by going deep into Model Context Protocol (MCP) and a problem that appears when you're connecting agents to something completely outside their training data: how do you teach an agent a language it's never seen before?
Expect technical detail, lessons from real implementations, and ideas you can apply to the AI systems you're building today.
David White - Do You Speak Press? Agent Language Lessons via MCP
Papermill uses a bespoke document language called Press, designed specifically for turning AI-generated content into PDFs. There's just one problem: unlike HTML, Markdown or LaTeX, no foundation model has ever encountered Press in its training data.
Simply exposing an API to an agent isn't enough. The agent might know how to call the tool, but it has no idea how to construct a valid payload.
David will deep dive into how Papermill built its MCP server to actively teach agents how to use its language, enabling them to successfully generate Press and interact with the service.
Using the real implementation as a case study, we'll look at patterns that can be applied to anyone building MCP integrations for specialist APIs, internal platforms, proprietary formats or other systems that sit outside a model's training set.
You'll learn:
- Why exposing an API through MCP isn't always enough
- How to teach agents concepts and formats they haven't encountered during training
- How Papermill designed its MCP server around a bespoke language
- Patterns for making specialist tools understandable and usable by agents
- Lessons you can apply when building your own MCP servers
----------------------------------------
Who Should Attend:
Software engineers, AI engineers, architects, tech leads and engineering leaders building AI agents, assistants or AI-enabled products. Whether you're implementing MCP servers and agent tooling or responsible for the experience around those systems, both talks offer practical lessons for building AI that works beyond the demo.
Why It's Worth Attending:
Practically AI is about the engineering reality of building with AI — real implementations, real problems and lessons you can take back to work.
This month we'll go from the internals of teaching an agent to work with completely unfamiliar systems through MCP, to the human realities of designing AI that people can actually use. Add pizza, questions and plenty of time to meet other engineers working with AI across ManchesterAgenda:
18:30 – Doors, pizza & mingling
18:50 – Introductions
19:00 – Talk: Do You Speak Press? Agent Language Lessons via MCP – David White
19:30 – Q&A
19:40 – Talk: TBC
20:10 – Q&A
20:20 – Open social & refreshments
21:00 – After event social at The Sawyer's Arms5 attendees
Past events
10
