Skip to content

Connecting Claude Desktop and ChatGPT to MCP servers

Claude Desktop اور ChatGPT کو MCP servers سے جوڑنا

35 min read

Three ways to see it

  1. Claude Desktop. After installing Claude Desktop from claude.ai, the configuration file sits at a specific path on each operating system. On macOS it is ~/Library/Application Support/Claude/claude_desktop_config.json. On Windows it is %APPDATA%\Claude\claude_desktop_config.json. The file is JSON. Inside, you declare an mcpServers object. Each key is the name you give the server. Each value names the command that launches it, along with arguments and environment variables. A typical entry: 'fbr-ntn': { 'command': 'node', 'args': ['/Users/admin/mcp/fbr-ntn/dist/server.js'], 'env': { 'FBR_API_KEY': 'redacted' } }. Restart Claude Desktop and the new tool appears in the model's repertoire.

  2. ChatGPT. As of late 2025, ChatGPT supports MCP servers through the developer mode in the desktop app and through the Custom GPT actions surface, with the protocol bridge maturing throughout 2026. The exact configuration UI is moving faster than this lesson; check OpenAI's docs for current steps. The principle is identical: a configuration file or settings panel where you register the server's launch command and any required credentials, after which the GPT can invoke its tools mid-conversation. The lesson here is not the menu path but the discipline of treating the configuration as a controlled artefact.

  3. A Pakistani team's likely config will list four to ten servers. A search server (Brave, Tavily, Kagi). A file system server scoped to a specific work folder. An internal Postgres or SQL Server connector. A custom server for FBR NTN lookup. Maybe a Slack server scoped to one channel. Each server is named, each path documented, each credential stored in a way that survives the engineer who set it up. The config file becomes a small piece of infrastructure that needs version control, code review, and rollback like any other piece of production.

Quick check

Quick check: what makes modern AI different from a rule-based program?

The why-tree

Why-tree level one: why a config file rather than a UI? Because text files version, diff, code-review, and survive across teams. A UI is a moment in time. A file is a record across time. For anything you want to audit later, prefer the file.

Try this with Claude

AI-edge prompt to try with Claude or ChatGPT: 'I am the IT head at a 200-person Pakistani bank. Draft an MCP deployment standard for my team that covers approved servers, credential management, offboarding, and PDPA-2023 audit logging. Output one Markdown page in English with an Urdu executive summary at the top.' Have a security colleague critique the draft.

Sources

Sources and further reading. Anthropic, 'Claude Desktop and MCP' (docs.anthropic.com/en/docs/build-with-claude/mcp). MCP configuration reference (modelcontextprotocol.io/docs/clients). Official servers list (github.com/modelcontextprotocol/servers). OpenAI ChatGPT desktop release notes (openai.com/blog). PDPA-2023 controller obligations summary (na.gov.pk).