The Model Context Protocol: connecting AI applications to tools and data through a shared, open standard.
The problem the protocol solves, its three primitives, and when to use it instead of a plain API call.
Implementing tools and resources, with schemas, errors and transports that behave.
A server is code with your credentials: trust, scope, approval and audit.
Shared servers, configuration, distribution and keeping an estate manageable.
25 in this topic
How the Model Context Protocol is structured: the host application, the clients it creates and the servers that expose capabilities.
MCP 1 min read 10 Sep 2025 For people
The three main things an MCP server can offer, who controls each, and when to use which.
MCP 2 min read 9 Sep 2025 For people
How MCP messages travel: standard input and output for local servers, HTTP for remote ones, and how to choose.
MCP 1 min read 8 Sep 2025 For people
A walkthrough of the steps to create a simple MCP server that exposes a tool, and how to test it.
MCP 1 min read 7 Sep 2025 For people
How remote MCP servers authenticate users with OAuth, and good practice for tokens and scopes.
MCP 1 min read 6 Sep 2025 For people
The main risks of connecting AI applications to MCP servers — malicious servers, injection, over-broad access — and how to reduce them.
MCP 2 min read 5 Sep 2025 For people
How tool names, descriptions and schemas in MCP servers determine whether models use them correctly.
MCP 1 min read 4 Sep 2025 For people
How MCP resources expose documents and data by URI, including templates, subscriptions and good design.
MCP 1 min read 3 Sep 2025 For people
How coding agents use MCP servers to reach issue trackers, databases, documentation and other tools.
MCP 1 min read 2 Sep 2025 For people
Common ways to structure MCP servers: API wrappers, workflow tools, read-only data access and gateways.
MCP 1 min read 1 Sep 2025 For people
Tools and techniques for checking an MCP server works: inspectors, logging, schema checks and model testing.
MCP 1 min read 31 Aug 2025 For people
How MCP relates to model function calling, framework plugins and direct API integration.
MCP 1 min read 30 Aug 2025 For people
Running MCP servers as hosted services: hosting, authorisation, scaling, monitoring and multi-tenant concerns.
MCP 1 min read 29 Aug 2025 For people
Client features that let MCP servers ask the host for a model completion or request information from the user.
MCP 1 min read 28 Aug 2025 For people
How organisations use MCP to connect AI assistants to internal systems with governance, security and scale.
MCP 1 min read 27 Aug 2025 For people
Frequent errors when building MCP servers — too many tools, stdout logging, oversized outputs, weak auth — and fixes.
MCP 1 min read 26 Aug 2025 For people
What happens between connecting MCP servers and a model calling a tool, and how hosts manage large tool sets.
MCP 1 min read 25 Aug 2025 For people
How MCP servers offer prompt templates that users can invoke, and good uses for them.
MCP 1 min read 24 Aug 2025 For people
An overview of MCP's ecosystem: SDKs, hosts, server registries, and how to assess third-party servers.
MCP 1 min read 23 Aug 2025 For people
How to turn an existing REST API into a useful MCP server without simply mirroring every endpoint.
MCP 1 min read 22 Aug 2025 For people