Your stack, in sync —
Connect Alphacast to your tools.
Bring your own sources in, and take the whole catalog out — into Excel, your models, your BI and your AI agents. One connection each way, and no scraper to keep alive.
Your files
Excel, CSV, PDFs
SharePoint & OneDrive
Your API
An internal endpoint
or a paywalled feed
Public sources
Central banks & agencies
queried live, not copied
Alphacast
One maintained catalog — cleaned, dated and cited.
Agents
Claude, Cursor, yours
27 MCP tools
Code
Python, R, notebooks
SDK and REST API
Spreadsheets
Excel and Power BI
Refresh, not re-download
Where the data ends up
Anything that can read a REST API, a Power Query connection or an MCP server — no new tool for your team to learn.
MCP: your AI builds on it, not just reads it.
The Model Context Protocol is how an agent reaches a tool. Point Claude, Cursor, a ChatGPT agent or your own internal assistant at the Alphacast MCP server. Catalog search, series, charts and pipelines are exposed as 27 tools, so the agent finds the data, builds the transformation, and saves the result somewhere it can keep refreshing.
MCP docs# from your own agent, over MCP
“Build me a chart of Brazil and Mexico CPI since 2020, and keep it current.”
- 1Discover — it finds the right series across every connected source.
- 2Build — the pipeline and the chart, from the prompt.
- 3Persist — saved into your workspace, not lost in the chat log.
What it leaves behind is an object, not a message: it redraws when the source publishes.
Python SDK
!pip install alphacast
from alphacast import Alphacast
alphacast = Alphacast("YOUR_KEY")
dataset = alphacast.datasets.dataset(38245)
df = dataset.download_data(format = "pandas", startDate=None, endDate=None, filterVariables = [], filterEntities = {})
REST API
GET https://api.alphacast.io/datasets/38245/data
→ your API key as the Basic Auth user.
The source and the last-updated stamp ride along.
The API and the SDK.
A dataset into a dataframe in two lines, filtered by variable or entity if you need it. The SDK reads and writes, so a notebook can publish back the series it just built — and every response carries its source and its last-updated stamp.
API and MCP are channels, not tiers — the same limits apply whether you pull from the web, a script or an agent.
Excel, and from then on it is Refresh.
Point Excel at the API once through Power Query and the series land as a table. After that it is Data → Refresh All, not a re-download — revisions included, pointed at the models you already have.
Bring your sources in.
The catalog is the starting point, not the boundary. Whatever you connect sits beside it and reaches the AI, the API and MCP the same way.
Excel, CSV and your own API
Upload a workbook or point us at an endpoint. It lands as a dataset like any other, with its own pipeline behind it.
SharePoint, OneDrive and PDFs
Point at the folder and the documents come in as structure — tables and series the AI can read, not attachments.
Hundreds of public APIs
Central banks, statistical agencies, market data. Not a copy: you query them live, use them as pipeline inputs, or reach them through MCP.
Custom connectors
A private provider, a paywalled feed, one of your internal systems. Our team builds the connector and keeps it alive when the provider changes it.
Your own data never counts toward download limits. What you bring is yours; the quota only ever applies to the Alphacast catalog.