Practitioner writing on MCP architecture, context engineering, and AI systems that survive contact with enterprise data. By the author of The Missing Layer.
One team, ten repositories, four tech stacks, and a Jira licence nobody enjoyed. We moved every ticket into markdown files next to the code. Seven months later, 15 projects and 12 people are on a board that costs nothing per seat.
Once tickets were markdown in the repo, the format turned out to be the easy part. The real deliverable was the process, written down in a form the agent executes: one skill, two write paths, and a loop you can tell not to stop.
Every ticket passes three automatic review layers before a human reads a line of it: a second model on the plan, a multi-lens self-review before the PR, and an agent-driven bot review after. The earliest one is the one that matters.
Eleven production MCP servers, no platform team, no framework dependencies, no API bill. The architecture that puts enterprise AI inside a mid-market budget.
Eleven production MCP servers, one mid-sized firm. No agent framework, no RAG pipeline, no AI-platform vendor. MCP plus enterprise identity is the whole stack.
Six fixes the MCP spec needs after 90+ tools and eleven APIs in production: from Resources no client surfaces to enums that break agent reasoning.
Three calls to the same tool with tightening filters. Without queryIntent: opaque retries. With it: the exact metadata gap, fixable in minutes.
An analysis of 856 tools across 103 MCP servers found 97.1% have at least one description smell. After shipping 90+ tools in production, here's the eight-block pattern that fixed ours.
Not all MCP servers are the same. Here's how the types differ — a six-level maturity ladder from hollow API wrappers to apps that write back, drawn from eleven production servers and 90+ tools. Where does yours sit?
Enterprise AI pilots fail 80% of the time. Surveys blame data quality. After eleven production MCP servers, the real problem was almost never the data.
Eleven production MCP servers at one mid-market firm. The complete framework: from understanding data through identity-bound deployment, end to end.