Ask TapeOps
Natural-language queries against your live book. “Top 5 losing clients as a share of B-book revenue today.” “Net USD exposure across EUR-crosses now.”
We use frontier models where they clearly help a human — queries, narration, drafting. We don’t ship chat UIs in your dashboard for the sake of it. Every AI capability has to earn its place on the screen.
Eight capabilities across three tiers. Read-only where the stakes are low. Supervised where they’re not. Every write goes through the engine’s approval queue — AI never calls MT5 directly.
Natural-language queries against your live book. “Top 5 losing clients as a share of B-book revenue today.” “Net USD exposure across EUR-crosses now.”
The desk sees strange P/L. TapeOps writes the one-paragraph explanation, cites the positions and timestamps, and suggests the next action.
Prose to RuleGraph. “Freeze any account whose daily drawdown exceeds 8% while news is within 30 minutes.” Drafts the rule, you approve, the engine executes.
Multi-step actions orchestrated by a sandboxed model agent. Every action hits the approval queue before it touches your book. Not autonomous — supervised.
Broker statements, KYC packets, PowerBI screenshots, PDF compliance filings — extracted into structured data the engine can act on.
MT5 server logs at scale. Triaged, narrated, and clustered by root cause. The on-call reads a summary, not a firehose.
Nightly risk, P/L, exposure, and anomaly digest. Delivered to the inbox your CRO actually opens.
Drafts only — never sends. Client notices, breach letters, desk memos. Tone + compliance guardrails built-in; human review before dispatch.