01
Pilot
Read-only observation of the account and the firm’s existing risk response.
- OKX and Bybit observation paths
- No trade instructions from pilot clients
- Breach observation and timing report
- Useful before vendor replacement decisions
One risk core. Three ways to deploy it.
Observe first, enforce when required, and benchmark the full reaction path using the same underlying state, rules and timing model.
Operating modes
Choose the least intrusive path that answers the risk question.
Pilot is for diagnosis. Watch is for enforcement. Bench is for measurement and technical evidence. They share the same rule logic and event model.
02
Watch
Live or demo enforcement for configured challenge rules.
- Real-time state and equity evaluation
- Parallel flatten and cancel operations
- REST fallback with verification
- Persistent keep-flat behaviour
03
Bench
Controlled demo breaches to measure end-to-end reaction latency.
- Market and synthetic trigger options
- Run-by-run request and fill timing
- Distribution and tail-latency reporting
- Reproducible dataset output
Cut sequence
From event to verified flat state.
Risk logic is only as useful as the transition from detection to exchange-confirmed outcome.
Ingest
Private and public exchange streams update the account state and current mark prices.
Evaluate
Equity is re-marked and configured rules are evaluated on relevant events and timed checks.
Enforce
Reduce-only closes and order cancellations are dispatched, with REST fallback where required.
Verify
The engine waits for exchange truth, records flat confirmation and keeps the account latched after breach.
Current exchange coverage
Clear about what is implemented now.
The site deliberately separates active enforcement support from read-only pilot support instead of implying identical coverage across every venue.
| Venue | Pilot | Active watch | Benchmark | Notes |
|---|---|---|---|---|
| OKX v5 | Supported | Supported | Supported | Private WebSockets, mark prices, close/cancel paths and demo benchmark tooling. |
| Bybit v5 | Supported | Not represented as production enforcement | Not represented as full benchmark mode | Read-only pilot path exists for observation and report generation. |
Engineering focus
Operational failure modes are product requirements.
Prop Rails has automated coverage around the cut path, retries, socket loss, rate limits, re-entry, persistence and timing behaviour.
Private WebSockets
Account, position and order streams keep risk state current.
REST fallback
Re-read and retry when the streaming path does not complete the cut.
Monotonic timing
Latency intervals use the host’s monotonic high-resolution clock.
Run evidence
Structured SQLite events support reproducible post-run analysis.
Use the pilot to establish your baseline.
Measure the current cut-off process before deciding how Prop Rails should fit into production.