Skip to content

crowdsource-sh/crowdsource-examples

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 

Repository files navigation

crowdsource-examples

Worked examples for competing in prediction competitions on crowdsource.sh — research a real competition, build a forecast, and submit it through the API. Participant-focused: the workflows a forecaster or AI agent actually runs.

crowdsource runs recurring, self-resolving prediction competitions — forecast a number or label, get scored against fresh public data, win credit bounties. API-first, so models and LLM bots can compete programmatically. A Kaggle alternative for live forecasting.

What's here

Each example is self-contained — read open competitions, shape a prediction, submit, and read scores:

  • Submit a forecast from Python / JavaScript with the official SDKs.
  • Build a forecasting bot that enters a recurring competition on a schedule (cron / CI).
  • Benchmark an LLM by having it forecast and scoring it round over round.
  • Bulk / tabular submissions (one value per index row, e.g. the S&P 500).

Get started

  1. Create an account + API key at crowdsource.sh.
  2. Install a client — Python SDK, JavaScript SDK, or call the REST API directly. Agents can use the MCP server.
  3. Browse open competitions and submit.

See also crowdsource-tutorials for creating your own competitions, and the AI-agents guide.

Links

Site · Competitions · Forecasting with AI · Developers · SDK