What is MCP
The Model Context Protocol (MCP) is an open standard that lets AI agents — Claude, Codex, and others — connect to the business tools and services you use, so they can do real work on your data instead of only talking about it.
Think of MCP as a universal adapter: a service publishes what an agent can do with it once, and every MCP-capable agent can plug in.
How it works
When you ask your agent to do something with a connected service, this is what happens under the hood:
sequenceDiagram
actor U as You
participant A as Agent (MCP client)
participant M as MCP server
participant B as The service
U->>A: prompt
A->>M: tool call
M->>B: query
B-->>M: data
M-->>A: tool result
A-->>U: response
- You ask your agent a question or give it a task.
- The agent picks a matching tool the MCP server exposes.
- The server runs the request against the underlying service and returns the result.
- The agent turns the result into a plain-language answer for you.
What this means for you
- Stay where you work. No jumping between chat and admin dashboards — your agent handles the task in the same conversation.
- Live data, not paste. Your agent works on what's actually in the system right now, not the snippets you copy in.
- Not locked into one agent. MCP is an open standard, so the same service works across compatible agents — Claude, Codex, and more as they adopt it.
- You stay in control. Agents ask for your confirmation before any write action, so nothing changes on your account without your sign-off.
Next step
- See what AfterShip Channels MCP can do for your business.
Learn more
The full protocol spec and the list of compatible clients are at modelcontextprotocol.io.