easyMCP is a visual builder for Model Context Protocol servers. Wrap any REST endpoint, test it in chat, and ship a hosted MCP URL ready for Claude, Cursor and ChatGPT — without writing a single line of server code.
Works with every major MCP client
Hand-writing MCP server code. Designing JSON-RPC tool schemas. Handling auth headers, OAuth flows and per-user tokens. Standing up a long-running host. Debugging protocol-level errors with no logs. For most teams, exposing a single REST endpoint to Claude is a multi-day infrastructure project.
easyMCP turns your REST APIs into hosted MCP tools visually. Schemas, auth, OAuth and the JSON-RPC protocol layer are handled for you. Test every tool inline, deploy to a stable HTTPS URL in one click, and inspect every call in real-time logs. No SDK, no Dockerfile, no infra to maintain.
Point easyMCP at any REST endpoint. Inputs, outputs, headers, and auth are all configured in a visual editor — no SDK, no boilerplate.
Talk to your tools through a chat UI that shows the model's reasoning, the exact request it sent, and the JSON it got back.
One click gives you a hosted MCP URL. Paste it into Claude Desktop, Cursor, ChatGPT, or any other MCP client.
MCP — the Model Context Protocol — is an open standard from Anthropic that lets large language models discover and call external tools through a single, well-defined interface. Think of it as USB-C for AI assistants: one port, every tool.
Before MCP, every assistant defined its own tool format. With MCP, one server works with Claude, Cursor, ChatGPT, Windsurf and any other compatible client.
Read the full explainerPOST https://easymcp.eu/api/public/mcp/prj_x9k...
Content-Type: application/json
{
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"name": "get_weather",
"arguments": { "city": "Berlin" }
},
"id": 1
}Drag, drop, and wire tools to REST endpoints. JSON schemas are generated for you.
A built-in AI chat shows real tool calls, arguments, and responses for fast iteration.
Get a hosted, HTTPS-stable MCP URL — no servers, no Dockerfiles, no DevOps.
Use a shared API key, prompt each end user for their own token, or full OAuth.
Claude Desktop, Cursor, ChatGPT, Windsurf, Continue, OpenAI Agents — all supported.
Secrets are encrypted at rest and never sent to the model — only the upstream API sees them.
Edge-deployed runtime keeps latency low even on the free plan.
Every deploy is immutable; roll back instantly if a tool change breaks.
Inspect every tool call, with arguments, response, status, and latency.
You already shipped the hard part: a working REST API. easyMCP turns each endpoint into a typed AI tool the model can reason about, with descriptions, schemas, and human-readable errors that LLMs can actually use.
GET /v1/customers/:id
→ { "id": "cus_123", "email": "ada@example.com" }A REST endpoint your backend already serves.
tool: get_customer(customer_id: string)
description: "Look up a customer by their ID."
returns: { id, email, plan, created_at }A typed MCP tool any AI assistant can call by name.
Give your team a Claude Desktop that can actually create Jira tickets, query Postgres, and ship Linear issues — backed by your existing internal APIs.
Wrap your CI, deploy, and feature-flag APIs as MCP tools so Cursor's agent can ship code end-to-end without leaving the editor.
Expose your product's REST API as an MCP server and let ChatGPT power workflows for your users — no plugin store required.
Hobby plan covers everything you need to wrap a few public APIs (weather, GitHub, Notion) and use them across every assistant you run.
Connect Snowflake, BigQuery, or Postgres through read-only tools so your assistant can answer real business questions safely.
Ship per-user MCP tools that prompt each customer for their own credentials — perfect for B2B SaaS.
easyMCP is designed around the principle that secrets never reach the model. Credentials are encrypted at rest with a project-scoped key, injected only at the moment a tool calls the upstream API, and stripped from logs.
Read the security overviewMCP is an open protocol from Anthropic that lets AI assistants discover and call external tools through a standard interface. Once your tools speak MCP, they work with Claude, Cursor, ChatGPT, and other compatible clients.
No. The visual builder generates everything — JSON schemas, the MCP envelope, auth handling. You just describe your REST endpoints.
Any MCP-compatible client. That includes Claude Desktop, Cursor, ChatGPT (via custom connectors), Windsurf, Continue.dev, and OpenAI Agents.
Credentials are encrypted at rest with AES-GCM and only decrypted at request time to call the upstream API. They are never exposed to the model or written to logs.
Custom domains are on the roadmap. Today every project gets a stable easymcp.eu URL that you can safely share with MCP clients.
Yes — 1 MCP project, 5 tools, and 100 calls per day, forever. The free plan shows a small easyMCP badge in the dashboard; paid plans remove it.