No code. Connect ready-made steps or import your OpenAPI, test them live and publish a secure server every AI client can call.
Free plan. No credit card. No code required.
Demo: an AI client asks a question and calls a workflow tool named get_todo_summary. The Start, HTTP Request, Condition and Return steps run in order and the tool returns a JSON result that the agent uses in its answer.
Works with every MCP client
Name it, pick a subdomain and a security mode. Five quick steps, live endpoint at the end.
Import your OpenAPI file, describe the tool to AI or connect steps on a canvas. Prefer code? JavaScript is one click away.
Install in Cursor, VS Code or Claude Desktop with one click, or add it to Claude Code, Codex and Gemini CLI with one command.
Every tool gets a name, a description and typed parameters. You only decide what it does, step by step.
No coding skills needed. Drop steps on a canvas, connect them and fill in a few fields. MCPCore turns it into a live MCP tool.
Need full control? Switch to a VS Code style editor with the sandbox SDK for HTTP, databases and secrets.
const res = await sdk.http({
url: `${env.API}/${params.id}`,
});
return res.json();Paste your OpenAPI or Swagger link. MCPCore reads every endpoint and turns the ones you pick into ready-to-use tools, with typed parameters and authentication filled in.
Demo: an OpenAPI link is pasted into MCPCore. Five endpoints are listed, the three read-only ones are selected, and they become live MCP tools that Claude can call.
See how an AI actually uses your tools before your users do, and keep every change you make.
Chat with Claude or GPT using your own API key and watch which tools the model picks, with the exact parameters and results.
Every save is a version. Keep drafts away from live traffic, compare changes side by side and roll back in one click.
No Git or GitHub needed. Every version lives here.
Everything a production MCP server needs, set up for every server you create.
Public, Unlisted, API Key or OAuth 2.0. Switch any time without touching your tools.
Store credentials once and reference them as env.KEY. Values reach the AI or the logs only if a tool outputs them.
Every call with its parameters, result and duration. Errors come with AI-assisted debugging.
Request volume, error rate, latency, top tools and the AI clients that call each server.
Serve your MCP endpoint from your own domain with automatic SSL.
Cap tool calls per minute for each server and protect the systems behind your tools.
Install your server in Cursor, VS Code or Claude Desktop with one click, or add it to Claude Code, Codex and Gemini CLI with one command. The right URL and auth header are filled in for you.
[mcp_servers.demo_server]
url = "https://demo.mcpcore.io/mcp"One command: codex mcp add
Add the block to your Codex config and restart Codex CLI or the IDE extension.
Explore MCP with personal projects.
Go production-ready with more power.
Scale up with advanced monitoring.
Tailored for large-scale operations.
Need it on your own infrastructure? Self-hosted Enterprise deployments are available on request.Talk to us
All prices in USD. Upgrades apply immediately, downgrades at the end of the billing period.
Create a free account and have a live MCP endpoint before your coffee gets cold. Questions about Enterprise or self-hosting? Send us a message.
Live in minutes
No Docker, no servers, no certificates to manage.
Secure by default
HTTPS, auth modes, encrypted secrets and rate limits on every server.
Fully observable
Every call logged, every error traced, every trend charted.