Skip to main content

MCP connector for AI assistants

MCP (Model Context Protocol) is a way to give an AI assistant access to an external service. With the Interconnect-Solutions connector you can send messages and check their delivery right from a chat: "send the customer a verification code", "did yesterday's message arrive?", "how much is left on the account?".

URI: https://mcp.interconnect.solutions/mcp

The transport is Streamable HTTP: every call is a JSON-RPC request sent with the POST method. You do not need to run a server of your own — the connector is already live.

Authorization​

Every call carries your API key in the Authorization header as Bearer <key>. The key travels straight to the platform API: the connector stores neither keys nor your messages.

Client configuration​

{
"mcpServers": {
"sms": {
"type": "http",
"url": "https://mcp.interconnect.solutions/mcp",
"headers": {
"Authorization": "Bearer ВАШ_API_КЛЮЧ"
}
}
}
}

Works with any MCP-capable client: Claude Desktop and Claude Code, ChatGPT, Cursor, VS Code, n8n and others.

Check the connection​

curl -s https://mcp.interconnect.solutions/health
# {"ok":true,"domain":"interconnect.solutions","send_enabled":true}

send_enabled tells you whether sending is on for your domain. When it is off, the sending tools are not offered at all.

What the connector can do​

Eleven tools. Read tools run straight away; sending tools always ask for confirmation first.

First call​

Ask for the balance — it costs nothing and proves the key works.

{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "balance",
"arguments": {}
}
}

Response​

{
"jsonrpc": "2.0",
"id": 1,
"result": {
"content": [
{
"type": "text",
"text": "{\"amount\": 152.4, \"currency\": \"EUR\"}"
}
]
}
}

Why sending requires confirmation​

Sending spends money from your account, so it has two locks.

Refusal response​

A sending tool called without confirmation refuses and never touches the API.

{
"refused": true,
"reason": "Sending spends the client's money and needs an explicit confirmation",
"how_to": "Ask the person, and pass confirm: true only once they agree"
}

How this differs from the regular API​

In substance, not at all: the connector calls the same https://api.interconnect.solutions as any other integration of yours. The difference is who composes the request — your code, or an assistant you briefed in words. Sending goes through the asynchronous endpoint: the message is queued and the response carries an identifier, not a delivery status.

Limits​

The full parameter list is on the Connector tools page.