Connect the MCP server to your AI assistant

One address, no account and no key. Once it is added, the assistant calculates Sri Lankan vehicle import tax with the published rates instead of guessing from memory.

What this does

MCP (Model Context Protocol) is the standard way for an AI assistant to use an outside tool. Our server gives the assistant a calculator: it asks you for the engine size, fuel, age and CIF value, runs the same engine as this website, and answers with every tax line, the rate version it used and the legal sources.

Free and open. No sign-in. A free API key is only needed for heavy use.
Read-only. Every tool only reads. Nothing is stored about you; the assistant sends only the vehicle type, size, age and CIF value.
Same numbers everywhere. The result equals the website, the REST API and the embedded calculator for the same rate version.

Use this address wherever an assistant asks for a server URL:

Server address (Streamable HTTP, no authentication)

https://autoimports.webgen.live/mcp

Claude (web, desktop and mobile)

Open Settings → Connectors.
Choose Add custom connector.
Name it AutoImports and paste https://autoimports.webgen.live/mcp. Leave any sign-in options empty: the server needs none.
Save. In a new chat, open the tools menu and make sure AutoImports is on.

Custom connectors are available on every Claude plan; the Free plan can add one. On Team and Enterprise plans an owner adds it once under organisation connector settings. If you cannot find the menu, search the settings for "connectors": the name and place move from time to time.

Claude Code

Terminal

claude mcp add --transport http autoimports https://autoimports.webgen.live/mcp

Run /mcp inside Claude Code to check that autoimports is connected.

ChatGPT

ChatGPT can use a custom MCP server in developer mode, which is offered on some paid plans. On a plan without it, use a Custom GPT instead (see below).

Open Settings → Apps & Connectors → Advanced settings and switch on Developer mode.
Back in Connectors, choose Create. Name it AutoImports, paste https://autoimports.webgen.live/mcp and choose No authentication.
Start a chat, open the + menu, choose Developer mode and tick AutoImports.

Developer mode warns that you are adding an unverified connector. That is expected for any server that is not in the ChatGPT app directory. The tools only read.

Gemini CLI

Terminal

gemini mcp add --transport http autoimports https://autoimports.webgen.live/mcp

Or add it to ~/.gemini/settings.json by hand:

settings.json

{
  "mcpServers": {
    "autoimports": { "httpUrl": "https://autoimports.webgen.live/mcp" }
  }
}

The Gemini app and gemini.google.com support custom MCP servers only on some plans and in some countries. If yours does not, see the link method.

VS Code, Cursor and Windsurf

Add the server to the editor's MCP file, then reload. Each editor names the address field differently:

VS Code: .vscode/mcp.json (or run “MCP: Add Server” and choose HTTP)

{
  "servers": {
    "autoimports": { "type": "http", "url": "https://autoimports.webgen.live/mcp" }
  }
}

Cursor: .cursor/mcp.json

{
  "mcpServers": {
    "autoimports": { "url": "https://autoimports.webgen.live/mcp" }
  }
}

Windsurf: ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "autoimports": { "serverUrl": "https://autoimports.webgen.live/mcp" }
  }
}

Le Chat, Copilot Studio and others

Assistant	Where to add it
Le Chat (Mistral)	Intelligence → Connectors → Add Connector → Custom MCP Connector. Paste the address; no authentication.
Microsoft Copilot Studio	Agent → Tools → Add a tool → Model Context Protocol. Use the address as the server URL with no authentication.
Goose, Zed, Cline, Continue	Add a remote or "streamable HTTP" server with the address above.
A client that only runs local (stdio) servers	Bridge it with `mcp-remote`, shown below.

Bridge for stdio-only clients

{
  "mcpServers": {
    "autoimports": { "command": "npx", "args": ["-y", "mcp-remote", "https://autoimports.webgen.live/mcp"] }
  }
}

Try it

Ask in plain words. A good prompt gives the assistant what the calculation needs:

Example

I want to import a 2021 Toyota Aqua hybrid (1,500 cc) from Japan.
The CIF value is ¥1,600,000. How much import tax will I pay in Sri Lanka?

Another

What does a Nissan Leaf (40 kWh, 2 years old) cost to import? CIF is £6,500.
Show every tax and tell me which rates you used.

You should see the assistant call the tool, then answer with:

each tax in order, then the total tax and the total landed cost in rupees;
the rate version and the date it was published;
a link back to the calculator for the same vehicle type.

If the assistant answers without calling a tool, say "use the AutoImports tool". If it asks you for the engine size, age or CIF value, that is correct: they change the tax.

The tools

Tool	What it does
`find_tax_class`	Suggests the tax class from a description such as “Toyota Aqua” or “electric van”. Returns up to three candidates and says what to ask when unsure.
`calculate_import_tax`	The calculation: every tax, the total landed cost, the rate version, the legal sources and a calculator link.
`list_tax_classes`	Every class, what it needs measured (cc, kW, seats…) and whether age matters.
`list_vehicle_categories`	The vehicle categories that are covered.
`get_applicable_rates`	Every excise and age band of one class.
`get_exchange_rate`	The LKR exchange rate for a currency, and where it came from.

There is also a resource, autoimports://rates/current: the live rate version and the legal sources. Every tool is marked read-only, so assistants will not ask to confirm each call.

In assistants that support MCP Apps (support varies by assistant and plan), a calculation also appears as a small card with the tax table, the totals, the rate version and a link to the calculator. Elsewhere you simply get the same figures as text.

Check the server yourself

To see what the assistant sees, list the tools from a terminal, or open the official inspector and connect to the address with the Streamable HTTP transport.

List the tools

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

Interactive inspector

npx @modelcontextprotocol/inspector

Limits and keys

Without a key, each IP address may make 300 calls a minute and 20,000 a day. That is much looser than the REST API (30 a minute), because AI providers call from shared addresses. A typical chat uses two or three calls.

With a free API key the limits belong to you, not to your address. Clients that allow custom headers take it as X-API-Key:

Claude Code with a key

claude mcp add --transport http autoimports https://autoimports.webgen.live/mcp \
  --header "X-API-Key: YOUR_KEY"

The chat apps that only take an address cannot send a header; they use the open limits, which is almost always enough.

If something does not work

What you see	What to do
The assistant says it cannot reach the server	Open `https://autoimports.webgen.live/mcp` in a browser. A short “method not allowed” message means the server is up (it expects a POST from an assistant). Check the address has no trailing characters and uses `https`.
It asks you to sign in	Choose “No authentication” or leave the OAuth fields empty. The server has no sign-in.
It answers from memory	Ask it to use the AutoImports tool. Check the connector is switched on for this chat.
“Too many requests”	You hit the limit; wait the time it states, or ask for a free key.
The numbers differ from another site or from Customs	Ours name the rate version and the gazettes. Compare the CIF value, the exchange rate and the age first. See why Customs can give a different figure.

Estimates only: the final duty is assessed by Sri Lanka Customs. Assistants are asked to say so; please keep that line when you share a figure.