Browse
Every MCP server, listed as the publisher shipped it.
Official registry, GitHub, npm, PyPI. A listing page sends you to their URL or command. We do not replace that with a PlugMCP endpoint.
Loading
MCP servers for people who do not write servers
Plug any tool
into any agent.
Browse MCP servers and install them from the publisher. Or build and host your own on PlugMCP when you want keys and logs from us.
Works withClaudeCursorCodexChatGPTWindsurf
A
Acme Ops
3 servers in one bundle
Endpoint
https://mcp.plugmcp.ai/s/acme-opscopy
Add to CursorAdd to ClaudeAdd to ChatGPTAdd to VS CodeAdd to Retell
Live calls
p95 640 ms
- 12:04:01okcrm.find_contact212 ms
- 12:04:03okcrm.create_deal418 ms
- 12:04:09oknotion.search155 ms
- 12:04:12errstripe.list_invoices1290 ms
- 12:04:20okgmail.send640 ms
- 12:04:33okcrm.find_contact190 ms
- 12:04:01okcrm.find_contact212 ms
- 12:04:03okcrm.create_deal418 ms
- 12:04:09oknotion.search155 ms
- 12:04:12errstripe.list_invoices1290 ms
- 12:04:20okgmail.send640 ms
- 12:04:33okcrm.find_contact190 ms
Install
Add it in Cursor, Claude, or VS Code.
The publisher's remote URL or local command, plus a row of clients. If they need an API key, the page says so. Your agent stores it.
Host
Need keys and logs? Build on PlugMCP.
That is a server you create in the app. HTTP tools on our gateway, app tools on n8n. Separate from the public directory listings.
How it works
Four steps. No terminal.
The hard parts of MCP are hosting, auth, and finding out why a call failed. PlugMCP handles those. You handle the part only you know: what the tool should do.
- 01Describe the tool
Pick an API or an app. Say what the tool does in one sentence. Name the inputs.
- 02Test it in the playground
PlugMCP lists the tools exactly as an agent will see them. Call one. Fix the description until the agent gets it right.
- 03Publish
You get an endpoint, an API key, and a short link with one-click installs for Cursor, Claude, ChatGPT, VS Code, and voice agents.
- 04Watch it work
Every call lands in your dashboard. Share the server in the directory if you want others to use it.
Voice mode
The same server, tuned for a phone call.
Voice agents cannot wait four seconds or read a JSON blob out loud. Switch on voice mode and PlugMCP serves a trimmed tool list, short descriptions, a hard timeout with a spoken fallback, and results summarised to a sentence. Attach it to Retell, Vapi, ElevenLabs, or OpenAI Realtime in one click.
# voice endpoint
https://mcp.plugmcp.ai/v/acme-ops
# policy
max_tools: 8
timeout_ms: 2500
on_timeout: "One moment, I am still checking."
result: summarise_to_sentence
Directory
Every MCP server, in one catalog.
Official registry servers sit next to GitHub and npm. Each listing points at the publisher. Skills live in their own catalog.
Guides
Install an MCP server yourself.
A listing page gives you the publisher's URL or local command, then a row of clients. These guides are the same steps, written out: Cursor, Claude, and what remote versus local actually means.
- 01
Add an MCP server to Cursor
Open the listing, click Add to Cursor, or paste mcp.json. What the agent sees after you allow the server.
Read the guide - 02
Add an MCP server to Claude
Claude Code is one terminal command. Claude Desktop uses a JSON file and mcp-remote for HTTP servers.
Read the guide - 03
Remote URL vs a local MCP command
A remote listing is an HTTP endpoint. A local listing is a command your agent starts. How to tell which one you have.
Read the guide
Your tools. Their agents. One cable.
Free plan: 2 servers, 1,000 calls a month, directory listing.