What is a MCP
MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems.
Think of MCP like a USB-C port for AI applications.
| Aspect | Description | Example |
|---|---|---|
| What it is | A server that exposes tools/data to the AI | Filesystem, GitHub, Jira, Postgres, Slack |
| What it replaces | Custom tool integrations per model | One MCP server works with any MCP-compatible model |
| How it works | Server exposes resources + tools; agent calls them | Agent reads a Jira ticket, then edits a file |
| Where configured | .mcp.json or IDE settings | Varies by harness |
When to use MCP: when the agent needs to read or write data from an external system (databases, issue trackers, APIs).
How MCP works
MCP uses a client-server architecture over JSON-RPC:
- The MCP client fetches a tool manifest that describes each available tool and when it should be used
- When the agent determines a tool matches the needs of its prompt, it calls that tool
- The MCP server executes the request and returns enriched context to the client
What an MCP server exposes
| Primitive | What it is | Example |
|---|---|---|
| Tools | Callable functions the agent can invoke | search_docs(), run_query() |
| Resources | Read-only data the agent can access | Files, database rows, API responses |
| Prompts | Pre-built prompt templates | Summarization, code review prompts |
How to use
Claude Code
claude mcp add --transport http notion https://mcp.notion.com/mcpOR
Create
.mcp.jsonand edit following the instructions
Copilot CLI
/mcp addOR
Create a
mcp-config.jsonin the.copilot folderfollowing the instructions
Types of MCP
| Transport | Description |
|---|---|
| HTTP | Remote MCP servers accessed over HTTP (recommended for cloud services) |
| stdio | Local MCP servers running as processes on your machine |
Tips
- MCP tool manifests consume context, so consider if you really need them in a certain context.
- For efficiency and simplicity, consider instructing the agent to use the CLI (e.g., GitHub CLI) directly or via a
skill.
Remember: Always use a reputable source for your
mcp.
Useful MCPs
| Category | MCP Server | Description |
|---|---|---|
| Frontend | Angular CLI MCP | Angular development tools |
| Frontend | Playwright MCP | Browser automation and testing |
| Frontend | Chrome DevTools | Chrome debugging tools |
| Frontend | Figma MCP | Design system integration |
| General | Context7 | Context management |
| General | Atlassian MCP | Jira, Confluence integration |
| General | GitHub MCP | GitHub repository management |
| DevOps | AWS MCP Servers | AWS service integration |
| DevOps | Azure MCP | Azure service integration |
| DevOps | Terraform MCP | Infrastructure as code |