Skip to content

The MCP server

Your workspace is also a remote Model Context Protocol server. An assistant that speaks MCP (Claude Code, Claude Desktop, Cursor and others) can read it for you: ask “what is driving the energy outlook this week?” and it calls your workspace’s tools to answer, from your own sectors and states.

Endpoint https://dash.alethman.com/api/mcp
Transport Streamable HTTP, stateless: one JSON-RPC message per POST, one JSON answer. No session, no event stream.
Authentication The same API key as the REST API, as Authorization: Bearer alm_live_… (Authentication)
Tools Eleven, all read-only (Tools)
Limits The key’s 120 requests a minute; every MCP message counts as one
Protocol versions 2025-06-18, 2025-03-26, 2024-11-05

Each tool runs the same code as a REST endpoint, under the same scope, and returns that endpoint’s JSON as text. So everything in the API reference (fields, limits, errors) holds for the tools too, with two differences made for conversations:

  • Smaller default pages. Where you name no limit, a tool asks for fewer items than the endpoint would (25 articles instead of 50, 25 forecasts instead of 100, 10 coordination findings instead of 30).
  • Answers that fit. No result is longer than 40,000 characters. A longer list is cut from the end with a truncated note (and, for articles, a cursor to carry on), and a long report is read a page at a time.

A key’s scopes decide which tools an assistant sees. Give the assistant’s key intel:read alone, and it can read stories, outlooks, forecasts and coordination findings, and nothing else.

The client opens with initialize; the server names itself and tells the assistant how to use the tools:

initialize → result
{
"protocolVersion": "2025-06-18",
"capabilities": {
"tools": {
"listChanged": false
}
},
"serverInfo": {
"name": "alethman",
"title": "Alethman",
"version": "1.0.0"
},
"instructions": "Alethman reads the news for this workspace, which follows 19 sectors. Every tool is read-only and returns JSON. Start with get_workspace to see what the workspace follows; list_outlooks for the current reading by sector; list_stories and get_story for events; list_signals for the articles themselves; list_reports and get_report for compiled reports. Ids from one tool feed the next."
}

Then the assistant calls tools as the conversation needs them:

tools/call
{
"jsonrpc": "2.0",
"id": 3,
"method": "tools/call",
"params": {
"name": "list_sources",
"arguments": {}
}
}
result
{
"jsonrpc": "2.0",
"id": 3,
"result": {
"content": [
{
"type": "text",
"text": "{\"data\":[{\"id\":\"34c8efbf-5e0c-4ace-bfeb-816f16bf59df\",\"url\":\"https://www.rbi.org.in/pressreleases_rss.xml\",\"site_url\":\"http://www.rbi.org.in/\",\"title\":\"PRESS RELEASES FROM RBI\",\"kind\":\"rss\",\"status\":\"active\",\"last_polled_at\":\"2026-11-21T05:11:08.842Z\",\"last_ok_at\":\"2026-11-21T05:11:08.842Z\",\"last_error\":null,\"items\":300,\"items_7d\":41,\"created_at\":\"2026-09-30T06:58:18.672Z\"}]}"
}
]
}
}

The tools read the narrative intelligence your workspace runs on: articles, stories, sector outlooks, forecasts, coordination findings (Veritos’s campaign view), reports and your sources. Veritos’s claim checks, Polity and Nostradamus are not offered over MCP or the REST API yet.