One limit framework across every venue, client and product you face — and something that holds it, live. Positions stream in as they fill, the whole book revalues against your limits, and every screen is pushed the result. Safetifi checks a ticket before it fills and turns a breach into an event with an owner and a deadline, instead of a number that turns red in tomorrow’s report.
Built for digital-asset prime brokers, OTC desks, brokers and exchanges.
| Client | Product | Cleared on | Client IM | Venue IM | Spread | Status |
|---|---|---|---|---|---|---|
Meridian Capital TIER-1 · BTC-PERP | BTC-PERP | Venue A | 12.00% | 8.50% | +3.50 | Pass |
Northwind Trading TIER-2 · ETH-PERP | ETH-PERP | Venue B | 10.00% | 9.25% | +0.75 | Thin |
Halcyon Markets TIER-1 · SOL-PERP | SOL-PERP | Venue A | 15.00% | 18.00% | −3.00 | Inverted |
Arbor Digital TIER-3 · BTC-0927 | BTC-0927 | Venue C | 20.00% | 14.00% | +6.00 | Pass |
Castlepoint LP TIER-2 · ETH-0927 | ETH-0927 | Venue B | 11.50% | 11.75% | −0.25 | Inverted |
A fill lands over a persistent socket with sequence tracking and resync, the book applies it as an event, the engine revalues exposure, margin, collateral and limits across the whole book, and the result is pushed to every open screen. No refresh button. No overnight batch.
An intermediary’s risk is spread across venues, clients and products that each enforce their own rules and none of yours. Every venue holds a line against you. Nothing holds your line across all of them.
Each venue enforces its own margin and position caps against you, in its own units, on its own schedule. The firm-level limit that spans them is a number somebody checks on a Monday — not a control that acts when it is crossed.
Post-trade reporting tells you that a limit broke. It cannot tell you that a ticket would break one. The only moment a limit is cheap to hold is before the fill, and that is the moment most systems have nothing to say.
When a venue raises initial margin on a symbol, every client tier priced beneath it silently becomes loss-making. Nothing in an order-management system flags it. You find out in the funding P&L, weeks later.
Under portfolio or unified margin, venue requirements move non-linearly with the book. A buffer that is comfortable at today’s prices can be gone on a 15% move — which is exactly when you can least afford to discover it.
Credit, margin, risk, collateral, lending and financing are not separate problems for an intermediary — they are views of the same exposure. Safetifi computes them from a single event-sourced book, so the numbers agree wherever you look at them.
Price client margin against venue margin, continuously.
The whole book revalued as it moves — measure the loss before it happens, not after.
What you give clients versus what you get from venues.
Set the limit before the client trades — and know which cap binds.
A term loan is not a margined position, and is not modelled as one.
The other half of the intermediary P&L.
Fees are a spread business too — what you bill against what you pay.
Answer the question while it is still a question.
One reconciled book, event-sourced end to end.
Alerts, stress tests and an audit trail that stands up.
Every desk watches something different. Let them build it.
Every screen answers the same question from a different angle: given both sides of this position, are we being paid enough to carry it?
The validator resolves the margin each client is charged — tier, product override, per-client override — and puts it side by side with the requirement of the counterparty holding that risk. Rows are ranked by how close they are to inverting.
Push positions, trades, clients and configuration over REST, WebSocket, FIX, NATS or SFTP — or let Safetifi pull from your venue accounts. Per-source authentication, schema validation at the boundary, corrections handled as events.
The engine resolves client margin, venue margin, collateral haircuts and financing rates on every update, revalues the book when prices move, and streams the result. A separate aligned UTC grid records the series to history. All money math is exact-decimal by construction.
Hold the line. Limits that act when they are crossed, a pre-trade check before the fill, alerts that carry an owner and a deadline, and new margin and rate schedules published with a full audit trail behind them.
Safetifi is an intelligence layer, not a migration. It reads from the systems you have — over the transport your operations team already supports — and stores state in your own database, behind your own identity provider.
Per-source credentials, replayable, corrections as events.
Venue directory with margin schedules, fees and connection state. Market data arrives as a signed snapshot — your deployment opens no exchange connections of its own.
One engine, deliberately. TimescaleDB accelerates the time-series history where it is available; without it the same schema runs on plain PostgreSQL.
Single sign-on against the directory you already run.
You extend leverage to clients and consume it from venues. Safetifi is the ledger of that difference — by client, by product, by venue, with the financing leg attached.
Bilateral margin terms negotiated per counterparty, cleared across several venues. Keep the net book, the basis and the collateral spread visible while the terms differ on every line.
Publish client-facing margin and financing schedules with confidence that no tier prices beneath the requirement you carry, and evidence the decision afterwards.
A term loan fails differently from a margined position: fixed principal, floating collateral, a real maturity and an LTV that runs the other way. Both books, side by side, with one valuation underneath them.
See the true cost of the leverage you are being charged, compare it across venues, and stress the book before the market does it for you.
A margin system is only as credible as the arithmetic and the audit trail behind it. Both are design constraints in Safetifi, not features added later.
Safetifi runs inside your VPC or on infrastructure dedicated to your firm. Your positions, clients and margin terms are never pooled with another institution’s.
Every margin, haircut, rate and P&L figure is computed with arbitrary-precision decimals. A floating-point number in a calculation path is treated as a defect, not a rounding preference.
Positions, corrections and published schedules are appended, never overwritten. Any number on screen can be traced back to the event and the configuration rule that produced it.
Market data is collected once, centrally, and distributed to deployments as a signed snapshot verified with a public key. Your installation never appears at a venue under your name or your IP, and a compromise anywhere else cannot forge prices into your book.
Authenticate against your existing directory over LDAP, OIDC or SAML. Permissions gate each functional area, so risk, treasury and operations see only their own surface.
We come from traditional finance, where real-time risk management on an intermediary book is a solved discipline — limits that hold, margin that is reconciled against the clearing house, and a control framework a regulator can read. We have spent the years since working in digital assets, where the same seat exists and almost none of that machinery does.
Real-time risk management in traditional finance — prime brokerage, clearing and margin, where the requirement you post and the requirement you charge are reconciled continuously and the limit framework is a control, not a report.
The market moves faster and never closes, and it is fragmented across venues that each set their own margin, their own haircuts and their own fee ladder, and change them without notice. The same position is priced differently on every venue you face.
Because an intermediary in that market has to answer two questions at once, continuously: is what we offer our clients competitive, and is it still in line with what the market and our counterparties are actually charging us? Nothing on the desk answered both.
Competitive for your clients, and in line with the market. At the same time, on every position.
That is the whole product. A digital-asset intermediary that only watches one of those two numbers is either losing business to a cheaper competitor or losing money on a spread that has quietly inverted — and in a market this fast and this fragmented, both happen in the same week.
No. Safetifi is an intelligence and control layer. It reads your book and your venue terms, computes margin, risk, collateral and financing, and tells you what to do. Orders stay in your OMS and assets stay with your venues and custodians.
A venue shows you the margin it charges you. It has no view of what you charge your clients, and no view of the other venues you face. Safetifi is built from the intermediary’s seat: both sides of every position, across every venue, in one book.
Positions, clients and your margin configuration. Most firms start with a read-only feed over REST or SFTP against a subset of clients, see the spread on their real book, then widen the connection.
Yes — with the model stated explicitly. Isolated and cross margin are computed in full. Portfolio and unified are computed as a worst-case scenario scan across the book, and the interface labels them as such rather than presenting an approximation as venue-exact.
In your environment. Single-tenant, on your own PostgreSQL — with TimescaleDB where it is available, to accelerate the time-series history — behind your own identity provider. Nothing about the deployment requires your book to leave your infrastructure.
A scoped pilot: we connect one ingestion channel against a slice of your live book and report the spread on it. If the spread is already where you think it is, you have evidence. If it is not, you have found the number this is for.
A pilot connects one channel against a slice of your live book and reports the two-sided spread on it — client by client, product by product.