Observation-only is a design choice, not a policy toggle
Observation-only infrastructure means your commercial platform can read telemetry, compute diagnostics, and send alerts, but it cannot close trades, modify overrides, or execute inherited apply actions. It is not a marketing toggle—it is an architecture decision that limits attack surface and legal liability.
In algorithmic trading, the temptation to add a remote panic button is understandable. Yet every mutating route is a vector for human error, credential compromise, or deployment bugs. Observation-only forces risk actions to remain on the terminal under operator or authorized EA control.
Teams running MT5 fleets often underestimate how many inherited routes exist in a commercial monolith: position closes, mode changes, registry deletes, tactical adjustments. Observation middleware must block those routes centrally when ObservationOnly=true, returning consistent, auditable responses.
Observation does not mean passivity. You can have aggressive alerts, near-real-time dashboards, multi-account correlation, and compliance reports. What you cannot do is mutate trading state from the public API. That line protects both the SaaS provider and the trader trusting the service.
For prop firms and funded accounts, observation-only proves independence: the monitoring tool does not compete with the EA or intervene in execution. That reduces questions about hidden copy trading, artificial latency, or remote manipulation during evaluations.
Implementing observation-only requires automated tests. Each protected route needs a test confirming HTTP 403 in observation mode and expected behavior in non-commercial internal environments. Without tests, a refactor can reopen dangerous endpoints unnoticed.
Operators must understand the difference between alert and action. A high drawdown alert invites terminal review; it does not auto-close trades. Training teams on that distinction avoids false expectations and unnecessary support tickets.
Integration design is affected too. Telemetry agents should be unidirectional. Dashboards should use GET. Alert webhooks may be outbound, but must never accept return commands that alter accounts. Think of observation as certified state reading, not remote control.
Traceability improves because every blocked mutation attempt can be logged. That creates evidence the platform rejected disallowed actions—useful in security reviews and third-party audits.
Quantisentry keeps ObservationOnly=true in commercial V1. Sentinel Cloud exposes bootstrap, telemetry, and public diagnostics while commercial middleware blocks protected routes. It is the right balance for monitoring SaaS before future phases with explicitly authorized actions.
Why promises are not enough
Fail-closed in code, not in documentation
Promising not to trade is easy; proving it in code is what matters. Centralized middleware evaluating each request against protected routes removes ambiguity. Clients can verify with standard tools that POST and DELETE to apply routes fail predictably.
Configuration validation should throw if someone tries to disable ObservationOnly in commercial builds. Defense in depth prevents accidental deployments with excessive permissions.
When to consider mutating actions
Only after explicit phases with strong auth, audit, WARN_ONLY modes, and operator approval should selective mutating actions open. Until then, pure observation reduces reputational and operational risk while you build trust in telemetry.
| Route type | Observation mode | Expected result |
|---|---|---|
| GET telemetry | Allowed | 200 with data |
| GET diagnostics | Allowed | 200 read-only |
| POST apply risk | Blocked | 403 fail-closed |
| DELETE registry | Blocked | 403 fail-closed |
FAQ
Does observation-only mean no alerts?
No. You can alert aggressively; only trading mutations from the API are blocked.
Can it be disabled in production?
Quantisentry commercial V1 keeps ObservationOnly=true; disabling requires an authorized phase.
How do I verify blocking?
Test protected routes with POST/DELETE; they should consistently return 403.
Does it affect EA performance?
No. The EA runs on MT5; the cloud layer only ingests telemetry via GET/agents.
Operational teams should document escalation paths when monitoring alerts fire during volatile market sessions.
Independent read-only monitoring complements broker dashboards with comparable metrics over time across accounts.
Structured telemetry makes post-trade reviews faster because every payload carries the same required fields.
Prop firm operators benefit from exportable evidence when challenge rules require proof of execution discipline.
Freshness SLAs turn silent expert downtime into visible incidents before capital is exposed to unmanaged risk.
Observation-only APIs reduce accidental execution risk because mutating routes remain blocked by design.
Risk committees prefer transparent scoring models where every penalty includes a written reason.
Multi-account operators should calibrate alert thresholds on a pilot fleet before scaling to production size.
Validation workflows must include out-of-sample tests and stress simulations before live deployment.
Documented contracts between EAs and monitoring platforms prevent ambiguous interpretations during audits.
Operational teams should document escalation paths when monitoring alerts fire during volatile market sessions.
Independent read-only monitoring complements broker dashboards with comparable metrics over time across accounts.
Structured telemetry makes post-trade reviews faster because every payload carries the same required fields.
Prop firm operators benefit from exportable evidence when challenge rules require proof of execution discipline.
Freshness SLAs turn silent expert downtime into visible incidents before capital is exposed to unmanaged risk.
Observation-only APIs reduce accidental execution risk because mutating routes remain blocked by design.
Risk committees prefer transparent scoring models where every penalty includes a written reason.
Multi-account operators should calibrate alert thresholds on a pilot fleet before scaling to production size.
Validation workflows must include out-of-sample tests and stress simulations before live deployment.
Documented contracts between EAs and monitoring platforms prevent ambiguous interpretations during audits.
Operational teams should document escalation paths when monitoring alerts fire during volatile market sessions.
Independent read-only monitoring complements broker dashboards with comparable metrics over time across accounts.
Structured telemetry makes post-trade reviews faster because every payload carries the same required fields.
Prop firm operators benefit from exportable evidence when challenge rules require proof of execution discipline.