Skip to content
Request access
The operating system for autonomous capital

Capital, intelligentlygoverned.

Connect research, risk, execution and oversight in one system.
You set the mandate. AI runs the firm.

Kernel · pre-trade check
Entry refused
Single-name concentration 7.3% vs 7.0% limit. Stored with the stage that refused it.
Trading policy
No single name above 7% of the account.
Largest name6.4% / 7.0%
Mandate v2.5Approved · 2 of 2
Decision record
Risk officer approved an intervention, with a written reason.
Second signature · 2 of 2
Adaptic OS · illustrative sample data

Institutional discipline. Without the institutional headcount.

A small team cannot check every order, watch every limit and record every decision by hand. Adaptic does it in software, every time, and brings people only what needs their judgment.

  • Limits that hold.
    Every order is checked against the account’s trading policy before it reaches the broker. A policy that cannot be read counts as trading switched off.
  • Protection that does not wait.
    Daily-loss and drawdown halts, trailing stops, volatility exits and the end-of-day flatten act on their own. Nobody has to be awake for risk to come down.
  • A record you can examine.
    Each trade carries the model’s thesis and the price that would prove it wrong. Each decision a person takes carries a written reason and, where required, a second signature.

Don’t add AI to the firm. Rebuild the firm around it.

Most firms bolt a model onto tools that were built for people. Adaptic starts from the other end. The trading call, the limits, the checks and the record live in one system, and people sit where judgment is needed.

JudgmentPeople
LimitsScreeningTrading callChecksBrokerRecord
  • Limits, set once.
    People write the limits. The engine works inside them every day.
  • Checks, not reminders.
    A limit is a number the engine tests before every order.
  • A record, not a recollection.
    Decisions are written down as they happen, with who and why.
  • Judgment where it matters.
    The system raises the exceptions. People decide them.
How it works

Author the mandate. Watch it run. Intervene with judgment.

Your people set the intent and the limits, watch the accounts, and decide what the system raises.

  1. Trading policy
    Single-name concentration
    ≤ 7.0%
    Sector concentration
    ≤ 25.0%
    Gross exposure
    ≤ 150%
    Open positions
    ≤ 20
    Entry halt · off · trading active
  2. Control Room
    Risk state · normal
    • Single-name concentration6.4% / 7.0%
    • Sector concentration19.8% / 25.0%
    • Gross exposure128% / 150%
    Entry refused · single-name 7.3% vs 7.0% limit
  3. Intervention
    Concentration escalation on one name.

    Raised with the evidence: the state, the expected impact, the downside, and what happens if nobody acts.

    • Portfolio manager · written reason✓ approved
    • Risk officer · second signature○ awaiting
    Each signature is recorded to the decision ledger.
Adaptic OS · illustrative sample data

Models are probabilistic. The permission is not.

A language model makes each buy, sell or pass call and must state the price that would prove it wrong. Everything after that is fixed rules: sizing, the checks, and the decision to send. If the model is unavailable, the trade is skipped. We call the checks the kernel.

  1. Screening
  2. Trading call
  3. Sizing
  4. Kernel
  5. Broker
  6. Record

Five places an order is checked

A staged risk check: account state, position count, single-name and sector concentration, exposure caps, margin, pattern-day-trader rules, daily loss, drawdown and buying power. An exposure the engine cannot measure blocks the open.

Illustrative · Refused · single-name concentration 7.3% vs 7.0% limit

Every refusal is stored with the stage that refused it and what it measured.

Each account moves through recorded risk states, normal, warn, throttle, paused, flatten and halt, and every transition is written down with its reason.

Humans govern. Machines operate.

The engine, on its own

  • Screening, the trading call, sizing and the checks
  • Exits, trailing stops and the end-of-day flatten
  • Halts on daily loss, drawdown, data quality and market dislocation
  • Broker reconciliation every five minutes in market hours

People, on the record

  • Set each account’s trading policy and limits
  • Stop new entries on an account with one switch; exits keep running
  • Write the mandate and approve it with a second person
  • Decide what the system raises, with a written reason and a second signature where required

Adaptic’s operators change engine settings only through a governed change that records the approver and the reason. Orders do not wait for a person.

Every trade explains itself.

Open any trade in the Adaptic OS and read why it exists: the signal behind it, the thesis, the invalidation and target levels, the legs, how far it moved for and against you, and its audit trail.

Ask about your book.

Ada, the OS’s AI assistant, answers questions about your book in plain language from live data through read-only tools. It cannot place or cancel orders.

Adaptic OS trade rationale: the thesis behind a trade, the signals it rested on and the conditions that would invalidate it.Adaptic OS trade rationale: the thesis behind a trade, the signals it rested on and the conditions that would invalidate it.
Adaptic OS trade rationale: the thesis behind a trade, the signals it rested on and the conditions that would invalidate it.Adaptic OS trade rationale: the thesis behind a trade, the signals it rested on and the conditions that would invalidate it.
A sample trade opened in Adaptic OS: the signal, the thesis, the target and invalidation levels, and the audit log.

Written as it happens.

Illustrative sample data
Written as it happens. · Illustrative sample data
WhenActorActControl
10:42:18Adaptic engineEntry refusedsingle-name concentration 7.3% vs 7.0% limit
11:05:02Adaptic enginePosition openedinvalidation 182.40 stated
13:20:47Adaptic engineTrailing stop movedposition stop
14:02:11Portfolio manager (sample)Intervention approvedrationale recorded, 1 of 2
14:09:36Risk officer (sample)Intervention co-signed2 of 2

Decisions are recorded as new entries, with who and why.

Who it serves

Built for the people accountable for capital.

Institutional governance without institutional headcount.

Set the limits once. The engine trades inside them, every order is checked before it reaches the broker, and the family’s advisers can read every decision.

What changes

  • Limits per account that the engine enforces
  • One switch that stops new entries
  • Decisions with a written reason and a second signature

Where we are building. Jurisdiction by jurisdiction.

Adaptic is based in the DIFC and is preparing to operate in ADGM, the United Kingdom, Singapore and the European Union. Each component and each jurisdiction shows its current state below.

  • 2

    separate grants, in two systems, before any account can send orders to a broker

  • 2

    people to approve a mandate; the author cannot approve their own

  • 5 min

    between broker reconciliations in market hours; a mismatch halts new entries

Structural facts about how the system is built. Not performance.

Status

  • Trading engine · US equities and two-leg options spreads, through the broker Running
  • Pre-trade checks, account scope and entry halt Running
  • Automatic protections · halts, trailing stops, end-of-day flatten Running
  • Decision records · risk verdicts, trade audit, governance decisions Running
  • Adaptic OS · Control Room, Interventions, Mandate Studio, book, risk, stress tests Running
  • Ada, the OS’s AI assistant · read-only Running
  • Order entry from the OS Switched off
  • Approved mandate driving the enforced limits Roadmap
  • Observatory screen for clients · trade lineage is recorded today Roadmap
  • NAV strike and fee accrual · preview today In build

Jurisdictions

  • DIFCNOW · LICENSING PENDING
  • ADGMIN PROGRESS
  • United KingdomIN PROGRESS
  • SingaporeIN PROGRESS
  • European UnionIN PROGRESS

A six-week shadow pilot. No capital at risk.

Your team works in its own Adaptic OS workspace on a broker paper account: your limits, your mandate, your approvers. Live market data, simulated execution.

  1. Weeks 1 to 2
    Set up.

    Invite your team, connect a paper account, set the trading policy, write and approve the mandate.

  2. Weeks 3 to 4
    Watch it run.

    The engine trades the paper account inside your limits. Your people watch, and decide what it raises.

  3. Weeks 5 to 6
    Review the record.

    Read the trades, the decisions and the refusals. Proceed, revise or decline.

  1. Weeks 1 to 2
    Set up.

    Invite your team, connect a paper account, set the trading policy, write and approve the mandate.

  2. Weeks 3 to 4
    Watch it run.

    The engine trades the paper account inside your limits. Your people watch, and decide what it raises.

  3. Weeks 5 to 6
    Review the record.

    Read the trades, the decisions and the refusals. Proceed, revise or decline.

Start with a conversation.

Tell us who you are and what you are evaluating. A person reads every request.

Sample data on this page is illustrative. Nothing here is an offer.