Free · Open source · Local-first

Chat with your AI. Sancho fetches the data.

Drop Sancho into Claude, Codex, or Cursor. Ask in plain English — it pulls real data from 120+ public sources (World Bank, FRED, Census, OECD…) and saves it to your repo.

Set up Sancho with a single message.

Install Sancho Fetch from PyPI (`uv tool install sancho-fetch`) and set it up on this computer. If `uv` isn't installed, tell me, then install it first. When you're done, tell me whether the install succeeded. Then explain in plain English, no jargon: what the .env file is, how to use it, where it is on my computer, and open it for me so I can paste API keys straight into it. Give me a few examples of data sources that need a free API key and a few examples of sources that don't need a key. Tell me where my downloaded data gets saved, and how I update Sancho later. Then run a short example data fetch using the World Bank so I can see where the data was saved. Finally tell me a few natural language examples I can tell you to fetch data so I can get a feel for how to use sancho-fetch and remind me that 1) I can use it anywhere as long as I'm in a LLM Desktop app's Code tab (e.g. ChatGPT Desktop or Claude Desktop), and 2) that I can ask AI to add new data sources by itself since sancho-fetch is built in a way that it's easily extensible (aka you can add more to it and the AI will do the heavy lifting and connect the new data source). As the team behind Sancho Fetch, we hope you truly make it yours as an internal resource to power your work.

Paste into Claude desktop → Code or ChatGPT desktop → Codex. That’s it! 🚀

  • cross air quality with childhood asthma, county by county
  • which federal grants reached counties this year?
  • build a climate vulnerability panel for 180 countries
Claude desktop → Code
ChatGPT desktop → Codex
open any folder
Step 01

Open your AI's code mode

Claude desktop → Code, or ChatGPT desktop → Codex. Pick any folder you like.

$uv tool install sancho-fetch
Installed sancho v0.1.0
$sancho setup
Step 02

Paste one message

Your agent installs Sancho from PyPI and sets it up. You watch.

U
/sancho fetch everything about population and demographics in Brazil since 2010
Fetched 1,284 rows from 6 sources → fetched-data/
Step 03

Chat to fetch

Type /sancho and ask in plain English. Works in any folder, anywhere on your machine.

sancho-workspace/
source/
custom/
my_module.py
playbooks/
Step 04

Extend, then update

Add your own modules and pull the latest sources whenever you want.

AI hallucinates data. Sancho fetches the real numbers from real sources.

122 sources, and growing

  • AHRQ MEPS
  • AHRQ NHQDR
  • AHRQ SDOH
  • EPA AirNow
  • ATSDR EJI
  • CDC/ATSDR SVI
  • ATUS
  • BEA NIPA
  • BLS
  • CDC BRFSS
  • CDC
  • CDC Biomonitoring
  • CDC Birth Defects
  • CDC Heat Events
  • CDC MMWR
  • CDC NHANES
  • CDC NHIS
  • CDC NSFG
  • CDC NVSS
  • CDC NWSS
How Sancho fits in

You chat. Sancho fetches. Files land in your folder.

You chat with your AI. The AI calls Sancho. Sancho pulls from the right module — one of 120+ curated, or one of yours — uses it to fetch the data, and drops it in your folder.

  1. 1
    You chat
    in any AI tool you already use
  2. 2
    Sancho fetches
    the AI calls the CLI for you
  3. 3
    From the right module
    curated, or one you built
  4. 4
    Saves to your repo
    real files you can open

Works with the AI you already use

Claude Code, Codex, Cursor, Claude Desktop, Antigravity, Perplexity Desktop — drop in, no new app to learn.

120+ data sources, growing

World Bank, FRED, Census, OECD, Eurostat — curated, audited, ready to fetch.

Self-host for your team

Run the MCP server in your stack and share Sancho with the whole org.

Charts, ready to render

A bundled skill file teaches your agent to turn fetched data into clean dashboards.

No tiers · No usage caps · No lock-in

Free and open source.

Sancho Fetch is open-source under Apache 2.0. Documentation is CC BY 4.0. Fork it, extend it, host it. The public repository is available now on GitHub.

Want help setting it up? Let us know — our team can help.

Open source

Want the repo? It’s on GitHub.

Browse the code, file issues, and contribute.

View on GitHub
Talk to a human

Interested in organizational deployment or building on top of Sancho?

Team setup, a missing data source, a contribution — we’d love to hear what you’re working on.

Send a message

Run it locally. Share it globally.

Fork it on your laptop. Stand up the MCP server for your team. Either way, it’s yours.