Model Context Protocol
Model Context Protocol (MCP) is an open standard that lets an AI agent connect to your tools and data (files, databases, CRMs, APIs) through one common plug, instead of a custom integration for every tool.
No jargon, no math
In plain words, what is it?
Before MCP, connecting an agent to a tool meant writing bespoke glue code for each pair: agent-to-email, agent-to-CRM, agent-to-your-ERP, hundreds of one-off adapters. MCP standardises the plug. A tool maker ships an MCP server that exposes what the tool can do (its actions, the inputs each one needs) in a common format. An MCP client built into the agent reads that description and can call the tool without bespoke code. You plug in a new system and the agent can use it straight away. Where it breaks: the server only exposes what its maker allowed, and the agent still needs permission to call it. MCP standardises access; it does not remove the need for guardrails.
- 01
Discover. The agent's MCP client finds the servers it is allowed to talk to.
- 02
List. Each server publishes the tools it exposes and the inputs they need.
- 03
Permit. The agent checks the call against its permissions; anything out of bounds goes to a human.
- 04
Call. The client sends a structured request to the server, no custom glue code.
- 05
Run. The server executes the action on the real system and returns the outcome.
Terms that live nearby.
Tool Use
Tool use is the capability of an AI model to do more than write text: it can call external tools, such as searching a database, sending an email or running a calculation, by emitting a structured request that software executes for real.
ReadFunction Calling
Function calling is the structured form of tool use: the model emits a machine-readable request naming a function and its arguments, written in JSON, which your code validates and executes, then feeds the return value back into the conversation.
ReadAI Agent
An AI agent is software that pursues a goal on your behalf: it reads the situation, plans the next step, acts through tools (email, calendars, databases, APIs) and checks the result before continuing, instead of answering one question and stopping.
Read§ 04
Want this running in your stack?
We design and ship AI systems with permissions, logs and human approval. It starts with a 30-minute scoping call.