API reference · Guides

MCP server

Your workspace as a Model Context Protocol server. Point any AI assistant or MCP client that speaks streamable HTTP, at it and ask about your conversations in plain language.

Connect

Endpoint https://icebot.icebergaisolutions.com/api/mcp, JSON-RPC 2.0 over HTTP POST, authenticated with the same API keys as the REST API. The key decides which agent the client sees.

Client configurationjson

{
  "mcpServers": {
    "icebot": {
      "type": "http",
      "url": "https://icebot.icebergaisolutions.com/api/mcp",
      "headers": { "Authorization": "Bearer ibk_XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX" }
    }
  }
}

Raw callshell

curl -X POST https://icebot.icebergaisolutions.com/api/mcp \
  -H "Authorization: Bearer $ICEBOT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Tools

A key is shown only the tools its scopes allow. With no scopes the server is read-only. Writes are internal and reversible: nothing here messages a customer, changes a script or deletes anything, and every write is recorded in the audit log against the key.

Tool	Scope	What it does
`list_conversations`	None (read)	Conversations, newest activity first. Arguments: limit, status (open, pending, closed).
`get_conversation`	None (read)	One conversation with its status, labels, AI summary and messages. Argument: conversation_id.
`conversation_report`	None (read)	Volume, how conversations finished, outcome mix and channel split for the recent period.
`set_conversation_labels`	`conversations:label`	Replace the labels on a conversation. Arguments: conversation_id, labels.
`set_conversation_status`	`conversations:status`	Move a conversation to open, pending or closed. Arguments: conversation_id, status.
`add_conversation_note`	`conversations:note`	Add an internal note, never sent to the customer. Arguments: conversation_id, body.

Methods: initialize, tools/list, tools/call and ping. See the endpoint reference for a full request and response.