Skip to main content

Glossary Term

MCP (Model Context Protocol)

By The Codegen Team · Updated March 26, 2026

An open protocol by Anthropic that standardizes how AI clients communicate with external services, tools, and data sources.

The Model Context Protocol (MCP) is an open protocol developed by Anthropic that standardizes how AI clients communicate with external services, tools, and data sources. MCP defines a structured way for AI assistants to discover available tools, understand their parameters, and invoke them through a consistent interface.

For coding agents, MCP enables integration with GitHub, Slack, Linear, Jira, databases, and custom internal tools without building bespoke connectors for each. An MCP-enabled agent can query a database, create a Jira ticket, and push code to GitHub using the same protocol.

MCP support varies across tools. Cline was one of the first to implement it as a first-class feature. Claude Code supports it natively. Cursor added MCP support in 2025.

In plain English

A standard connection layer that lets AI tools talk to your other software — GitHub, Slack, Jira, databases — without building a custom integration for each one.

Why it matters

Before MCP, connecting an AI agent to an external service meant writing bespoke integration code. MCP standardizes that so one protocol works across all supported tools. An agent can read a GitHub issue, check a Slack thread, and open a pull request without any of those connections requiring separate setup.

In practice

A developer asks their agent: "Check the Linear tickets tagged high-priority, fix the top three bugs, and post a summary to #eng-updates on Slack." With MCP connectors configured, the agent does exactly that — Linear, Slack, and GitHub are all accessible through the same protocol. Without MCP, that workflow requires three separate integrations or a lot of copy-pasting.

How Codegen uses MCP (Model Context Protocol)

Codegen supports MCP natively, so agents can reach GitHub, Slack, Linear, Jira, and custom tools without extra setup. The practical impact: a Codegen agent assigned a ClickUp task can pull in the PR from GitHub, read the related discussion in Slack, and update the Linear ticket status when it opens a pull request — all in one execution. What MCP does not solve is which external data is actually useful for a given task. Codegen handles that through task context, not just tool access.

Frequently Asked Questions