Guides
Sandbox & test data
Build and test against a full-fidelity environment with simulated rails, deterministic settlement, and on-demand scenario triggers.
What sandbox is
The AgentRail sandbox is a complete, isolated copy of the platform. The API surface is identical to production — same endpoints, same payloads, same error shapes. The only differences: no real money moves, and you can trigger any scenario on demand.
Funding test wallets
Sandbox wallets start with a zero balance. Credit them via the sandbox helper:
Python
await client.sandbox.wallets.credit(
wallet_id=wallet.id,
amount=1_000_000_00, # $1,000,000.00
)Simulating scenarios
- ─
sandbox.payments.simulate_failure— force a failure with a specific reason code. - ─
sandbox.compliance.simulate_hit— trigger a sanctions screening hit. - ─
sandbox.disputes.simulate_chargeback— file a dispute against a sandbox payment. - ─
sandbox.webhooks.replay— replay any past webhook for testing your handler.
Note
Scenario triggers are sandbox-only. They're rejected with a clear error if you accidentally point a sandbox key at production.
Isolation guarantees
Sandbox and production are physically separate. Sandbox keys begin with ar_test_; production keys begin with ar_live_. You can't accidentally move sandbox funds into production or vice versa — the platform refuses the call at the credential boundary.