MCP (Model Context Protocol) Without Code: Connect Claude to Your Apps and Data

Why MCP exists

Before MCP, every AI app needed a custom integration for every tool: M apps × N tools. MCP, an open standard introduced by Anthropic in November 2024 and now supported by Claude, ChatGPT, Gemini, Microsoft Copilot Studio, Cursor, VS Code and many others, works like USB-C for AI: build an MCP server for a tool once, and any MCP-capable assistant can use it.

The architecture in plain words

Using MCP in Claude without code

1. Built-in connectors (easiest)

In Claude (web or desktop): Settings → Connectors (or the "+" / tools menu in a chat). Connect Google Drive, Gmail, Google Calendar, GitHub, Slack, Notion, Asana, Atlassian (Jira/Confluence), Linear, Canva, Figma and others from the directory. Each connection uses OAuth, so you choose the account and can disconnect any time.

Try: - "From my Drive, find the Q3 roadmap and list the three riskiest items." - "Summarise unread emails from customers this week into a table with sentiment." - "Create Jira tickets for every bug in this meeting transcript, assigned to the component owners."

2. Custom remote MCP servers

Many companies publish a remote MCP server URL (for example https://mcp.example.com/mcp). In Claude: Settings → Connectors → Add custom connector, paste the URL, then sign in. Organisation admins can approve connectors for the whole team.

3. Local servers in Claude Desktop

Claude Desktop can run local MCP servers (files on your laptop, a local database). Many are installed as desktop extensions with one click; others are added in the app's developer settings with a small JSON entry that names the command to run. Ask your engineering team for the approved list.

Build MCP tools without code

# What a tool definition looks like to the model: a name, a description and an input schema (JSON Schema)
import json
tool = {"name": "create_lead", "description": "Add a sales lead to the CRM sheet. Use when the user shares a new prospect.",
        "input_schema": {"type": "object", "properties": {"name": {"type": "string"}, "email": {"type": "string"},
                         "company": {"type": "string"}, "source": {"type": "string", "enum": ["website", "event", "referral"]}},
                         "required": ["name", "email"]}}
print(json.dumps(tool, indent=1))
call = {"name": "create_lead", "arguments": {"name": "Ravi Mehta", "email": "[email protected]", "source": "event"}}
missing = [f for f in tool["input_schema"]["required"] if f not in call["arguments"]]
print("valid call" if not missing else f"missing {missing}")

{
 "name": "create_lead",
 "description": "Add a sales lead to the CRM sheet. Use when the user shares a new prospect.",
 "input_schema": {
  "type": "object",
  "properties": {
   "name": {
    "type": "string"
   },
   "email": {
    "type": "string"
   },
   "company": {
    "type": "string"
   },
   "source": {
    "type": "string",
    "enum": [
     "website",
     "event",
     "referral"
    ]
   }
  },
  "required": [
   "name",
   "email"
  ]
 }
}
valid call

The description is what the model reads to decide when to use the tool, so PMs often write and test these descriptions the way they write user stories.

Security and governance

Use cases for PMs, BAs and senior PMs

Summary

Learn more

• Model Context Protocol documentation
• Claude connectors help
• n8n MCP Server Trigger
• Zapier MCP