One MCP server, eight clients: the config differences that break the setup

The promise of MCP is that one server works in every client. It mostly does. What differs is the paperwork: the file the config goes in, the name of one key, the word for the transport, how long the client waits. Almost every "the server is added but I see no tools" report I have read comes down to one of a handful of those differences.
So here they are in one place, checked against each client's own documentation on 14 November 2026. To make it concrete, every example adds the same server: the free, read-only MCP server of Recipe for Agents, which I maintain. It needs no account and no key, so you can paste any snippet below and see the tools appear. Swap in your own URL afterwards.
The same URL, eight shapes
| Client | Where the config lives | What you write |
|---|---|---|
| Claude Code | command, or .mcp.json for project scope |
claude mcp add --transport http <name> <url> |
| Claude Desktop / claude.ai | Customize > Connectors > + Add > Add custom connector | name + URL in a form |
| Cursor | ~/.cursor/mcp.json or .cursor/mcp.json |
mcpServers > url |
| VS Code | .vscode/mcp.json |
servers > "type": "http" + url |
| Codex CLI | ~/.codex/config.toml |
[mcp_servers.<name>] + url |
| Gemini CLI | command, saved in settings.json |
gemini mcp add -t http <name> <url> (httpUrl in the file) |
| ChatGPT | Plugins > + > Add custom MCP server | URL + authentication in a form |
| n8n | MCP Client Tool node on an AI Agent | Server Transport: HTTP Streamable, Endpoint |
For the example server that means:
claude mcp add --transport http recipe-for-agents https://recipeforagents.com/mcp-server
gemini mcp add -t http recipe-for-agents https://recipeforagents.com/mcp-server
// Cursor: ~/.cursor/mcp.json
{ "mcpServers": { "recipe-for-agents": { "url": "https://recipeforagents.com/mcp-server" } } }
// VS Code: .vscode/mcp.json
{ "servers": { "recipe-for-agents": { "type": "http", "url": "https://recipeforagents.com/mcp-server" } } }
# Codex CLI: ~/.codex/config.toml
[mcp_servers.recipe_for_agents]
url = "https://recipeforagents.com/mcp-server"
Now the places where it goes wrong.
1. mcpServers in one editor, servers in the other
Cursor, Claude's .mcp.json and most READMEs use a top-level mcpServers key. VS Code's .vscode/mcp.json uses servers, and a remote entry also wants "type": "http". Paste a Cursor config into VS Code and nothing complains; the server just never shows up.
One more thing to know if you live in VS Code: its documentation now calls this setup the "Local harness" and says it is deprecated and will be removed in February 2027. The replacement is the Copilot format, which uses the portable mcpServers key in ~/.copilot/mcp-config.json or a workspace .mcp.json. So the odd one out is slowly coming back into line.
2. httpUrl and url mean different transports in Gemini CLI
In Gemini CLI's settings.json, httpUrl is a streamable HTTP endpoint and url is an SSE endpoint. Write url for a server that only speaks streamable HTTP and Gemini CLI tries the wrong protocol: the server is listed, the tools are not. Adding the server with gemini mcp add -t http writes the right key for you. Two smaller notes: without -s user the server goes into the project's .gemini/settings.json, not your global one, and a header you pass with -H (a bearer token, say) ends up in that file as plain text, so keep it out of version control.
3. Codex gives a tool call 60 seconds
Codex CLI's tool_timeout_sec defaults to 60 seconds, and startup_timeout_sec to 10. That is fine for a lookup and too short for anything that collects data for a minute or two: the call is abandoned while the work carries on at the other end. Raise it in the server's table:
[mcp_servers.my_slow_server]
url = "https://example.com/mcp"
bearer_token_env_var = "MY_SERVER_TOKEN"
tool_timeout_sec = 300
bearer_token_env_var names an environment variable instead of putting the token in the file, which is the habit worth keeping in every client that allows it.
4. Claude Code cuts long tool output
Claude Code warns when one tool result passes 10,000 tokens and stops at 25,000 by default. A server that returns a whole dataset in one answer hits that quickly. Start Claude Code with MAX_MCP_OUTPUT_TOKENS=50000 (or whatever you need), or better, ask the server for fewer rows. Inside a session, /mcp shows each server's status and runs the OAuth sign-in for servers that need one.
5. Claude's custom connectors call your server from Anthropic's cloud
This one surprises people who test against a local server. A custom connector in Claude Desktop or claude.ai is not reached from your computer: Anthropic's documentation says the connection comes from its cloud, so a server on localhost, on a private network or behind a VPN will not connect, even when your browser can open it. Use Claude Code (which runs on your machine) for local servers, or put the server somewhere public. The menu has also moved: it is Customize > Connectors > + Add > Add custom connector.
6. ChatGPT reads the tool list once
In ChatGPT the place is now called Plugins: + then "Add custom MCP server", a URL, an authentication choice, and "Create as a plugin". ChatGPT discovers the tools at that moment. Change a tool's name, description or schema on the server and ChatGPT keeps the old list until you open the connection under Plugins and press Refresh. Availability depends on your plan and workspace settings, so if the option is missing, that is the first thing to check.
7. n8n's docs and n8n's node disagree
n8n's documentation page for the MCP Client Tool node still describes a single "SSE Endpoint" field. The node itself has a Server Transport option with HTTP Streamable (the default in recent versions) and Server-Sent Events, plus an Endpoint field. If your node only offers SSE, update n8n before blaming the server. Authentication options are None, Bearer, Header, Multiple Headers and OAuth2.
8. Configs with /sse in the URL are living on borrowed time
The MCP specification replaced the old HTTP+SSE transport with streamable HTTP in its 2025-03-26 revision: one endpoint that takes POST requests. Servers are dropping the old endpoints, and a config that ends in /sse is usually the reason an old setup "suddenly" stopped working. Point it at the server's current URL with the streamable HTTP transport.
A two-minute check for any new server
- Add it with the client's own command where one exists (Claude Code, Gemini CLI): the command writes the right keys.
- Open the client's server list (
/mcpin Claude Code and Gemini CLI, the MCP Logs output panel in Cursor) and confirm the tools are listed, not just the server. - Call the cheapest tool first. On the example server, ask "which tool gets me LinkedIn company headcounts?" and it answers from its catalogue without running anything.
- Only then add secrets, and keep them in environment variables or the client's prompt-for-secret feature (VS Code's
inputswith"password": true) rather than in a file you might commit.
Limits of this page
Clients change their MCP support every few weeks, menus included. Everything above was checked against the official pages listed below on 14 November 2026, not against every client version in use. If something has moved, the vendor's page wins.
Sources: Claude Code MCP · Claude custom connectors · Cursor MCP · VS Code MCP configuration · Codex MCP · Gemini CLI MCP servers · ChatGPT: connect from ChatGPT · n8n MCP Client Tool · MCP transports, 2025-03-26
I run Recipe for Agents, which collects MCP configs, n8n workflows and agent skills for live web data; the example server is ours. Drafted with AI assistance and checked line by line against the sources above.
About the author
- Recipe for Agents collects copy-ready MCP configs, n8n workflows and agent skills for giving AI agents live web data: recipeforagents.com
Comments
Join the discussionNo comments yet. Be the first to comment.

