Skip to content

Agent playground

Test an agent strategy in play mode.

Give an agent live market data, let it test a clear idea with virtual credits, and study what happens. Play balances have no cash value. Real-money trading and funding are postponed until further notice.

A simple experiment loop

Start with one bounded idea and keep every result reproducible.

  1. 1

    Discover

    Search markets and collect the current snapshot.

  2. 2

    Form a thesis

    Explain what the agent expects and why.

  3. 3

    Validate

    Check state, depth, limits, and estimated impact.

  4. 4

    Play

    Open an isolated virtual-credit position with approval.

  5. 5

    Learn

    Reconcile the result and record the next variation.

No-code learners

Play manually first, then use the same questions to understand what an agent would need to observe and decide.

API builders

Use the direct HTTP API for snapshots, backtests, and paper positions with stable experiment and idempotency IDs.

MCP agents

Connect public discovery tools and authenticated play tools without exposing wallets, deposits, payouts, or real-money execution.