Agent
Environments

Agents act as people to predict behavior.

Simulations— real-world scenarios, ready to run Environments— build custom scenarios, quantify outcomes MCP— connect your agent, ask it what if questions
How it works
How it works

Agents in place of people.
Observe how they behave.

1 — build the envthe scenario
2 — drop in agentsas people
3 — watch what they dorun 1
Simulations

Real-world scenarios.
Ready to run.

tested against real outcomes · publications →

market
Will customers still buy if we raise prices?
exchange
What happens to the stock if earnings disappoint?
council
Cut, hold or hike? Experts debate it, then vote.
dispute
A disagreement? See who's right in a virtual court.
auctionsnegotiationssupply chainssocial mediaelectionslegislature+ more
Environments

Build custom scenarios.
Quantify outcomes.

1 — askan agent builds it for you
2 — the environmentwaiting…
price_cut_20.env market you -20% price competitor_a competitor_b consumers outcome: market_share @ 12mo
—
3 — parallel runslive
price_cut_20 — liveruns: 0 / 1000
— —
MCP

Connect your agent.
Ask it what if questions.

Connect agent →
claude · sessionfareground mcp
connect
Open source

The foundations. Open source.

Build on the SDKs →
environments-sdk
Rules, participants and outcomes,
written as one contract.
agents-sdk
Brains, tools and lifecycle for
the agents inside environments.