Preparing your Team for the Context Layer
A practical guide to getting your organisation ready for the context layer: the people, APIs, executive sponsorship, use cases, and domain knowledge you need to start delivering real value from AI agents.
Blog
Enterprise Context Challenges
Connecting AI to enterprise systems is easy; returning the right context is hard. A look at the identity, context engineering, and sprawl challenges that emerge beyond the demo, and principles for getting them under control.
APIs as Skills via MCP
Agents can now consume web APIs as Skills via MCP. Progressive disclosure and dynamic tool registration replace bloated tool preloading, with auth and access control enforced server-side.
Skills in MCP
Skills and MCP are not rivals. Skills are portable context for agents, and MCP is the protocol that lets you serve them remotely with auth, oversight, and access control.
What makes Skills good
How a simple file convention turned prompt folders into portable, discoverable Skills, and the challenges that come with them.
Beyond RAG - methods for context disclosure
Exploring pre-loaded context, RAG, and in-context agentic retrieval as methods for disclosing relevant resources to LLMs without overwhelming them.