MiniPACS + Vendo

Developers

MiniPACS MCP server

Connect Claude, ChatGPT, Cursor or any MCP client to minipacs.net and ask about MiniPACS and Vendo: prices, features, limits and deployment, answered from the published pages with their source links. Public, read-only, no key.

Updated November 2026

Endpoint

  • URL: https://minipacs.net/mcp
  • Transport: Streamable HTTP, MCP protocol 2025-06-18, stateless.
  • Auth: none.
  • Server card: /mcp/server-card; discovery file /.well-known/mcp.json; registry name net.minipacs/minipacs.

Tools

  • ask: retrieves the passages of minipacs.net that answer a question, with their URLs (NLWeb v0.55). Input: query.text (up to 2,000 characters together with the optional context) and prefer.limit (1 to 20, default 5). Read-only and idempotent.
  • send_feedback: tells the site's owner how a task on the site went. Input: outcome (success, partial or failure), and optionally task, detail and page. A person reads it; it changes nothing on the site.

Questions it answers well: what MiniPACS costs for two locations, whether Vendo has an HL7 interface, how backups work, what a self-hosted PACS needs, how MiniPACS compares with a cloud PACS.

Connect a client

Claude Code:

claude mcp add --transport http minipacs https://minipacs.net/mcp

Claude (claude.ai and Claude Desktop): Settings, Connectors, Add custom connector, then paste https://minipacs.net/mcp.

ChatGPT: in developer mode, Settings, Connectors, Create, then paste the same URL.

Cursor (~/.cursor/mcp.json):

{
  "mcpServers": {
    "minipacs": { "url": "https://minipacs.net/mcp" }
  }
}

VS Code (.vscode/mcp.json):

{
  "servers": {
    "minipacs": { "type": "http", "url": "https://minipacs.net/mcp" }
  }
}

Raw JSON-RPC

Any HTTP client works. List the tools, then call ask:

curl -s https://minipacs.net/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

curl -s https://minipacs.net/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"ask",
       "arguments":{"query":{"text":"Does MiniPACS charge per study?"},"prefer":{"limit":3}}}}'

The same search is a plain HTTP API too: GET https://minipacs.net/ask?query=..., described in /openapi.json.

Limits

  • It knows what minipacs.net publishes, nothing else. The neutral summary is in /llms.txt and Facts for AI search.
  • It cannot book a demo or contact anyone. In a browser, the WebMCP tool prepare_demo_request opens and fills the demo form for the person to send; the e-mail is [email protected].
  • It is not a way into a clinic's MiniPACS. Inside the product, order systems use the order-system API.

FAQ

What is the MiniPACS MCP server?

A public Model Context Protocol server at https://minipacs.net/mcp. It lets an AI assistant such as Claude, ChatGPT or Cursor search everything published on minipacs.net about MiniPACS (self-hosted PACS) and Vendo (referral portal) and answer with the source links: prices, features, limits, deployment and guides.

Does it need an API key or an account?

No. It is read-only, needs no key and keeps no session. It only reads the public pages of minipacs.net.

Can it see patient data or a clinic's PACS?

No. It reads the public website only. It has no connection to any MiniPACS or Vendo installation, and those run on each clinic's own server.

Which tools does it have?

Two: ask, which searches the published pages and returns passages with their URLs, and send_feedback, which tells the site's owner how a task went. Neither changes anything on the site.

Where is it listed?

In the official MCP Registry under the name net.minipacs/minipacs, in the AI catalog at https://minipacs.net/.well-known/ai-catalog.json, and in https://minipacs.net/.well-known/mcp.json.

Own your archive: the MiniPACS + Vendo brand card for independent imaging centers

See it run, then get pricing.

Built for US imaging clinics and proven in production: millions of images already live on this exact stack. Flat $300 a month per location, no per-study fees, and you can export everything to standard DICOM whenever you want. Both demos are the real apps on synthetic data; enter a work email and we send a one-time link that opens them.

Get pricing