Structured products. AI-native.

From client brief to review-ready term sheet.

OpenArrow takes structured products from brief to price, stress, document and review in under 60 seconds, powered by one deterministic engine. Work in chat, through your AI assistant, or via API.

See the workflow
Every figure engine-computedDaily-calibrated vol surfacesPer-key audit trail

Desk run

3Y NIFTY capital-protected note

Review-ready
  • 09:12

    Brief · RM, private wealth

    “3 years on NIFTY, full capital protection, best participation you can get. ₹10L ticket.”

  • 09:12

    Priced · arrowquant

    NIFTY vol surface, latest close · A1+ funding 9.50%

    Quote
  • 09:13

    Stressed · arrowquant

    Spot ±20% × vol ±5 pts · investor SRI 1 of 7

  • 09:13

    Documented · arrowquant

    KID and term sheet generated

    KID.pdf
  • 09:13

    Logged · usage dashboard

    Three engine calls recorded against key oa_k…7f2: tool, status, duration and time, linked to the chat turn.

NIFTY · Gold

Index and gold underlyings

< 60 s

Brief to review-ready KID

10

Payoff families

17

Templates on the Structures shelf

61

MCP tools for your AI assistant

The workflow

Five steps a desk already takes. Minus the spreadsheets between them.

Each step hands the next one the same numbers. The quote, the stress grid and the KID come from one engine run, so they agree.

01 · Brief

Start from the client’s words.

An RM or PM describes the need in plain language: in OpenArrow chat, inside the AI assistant the team already uses, or through the API. The agent maps it to a payoff family and asks for what is missing.

Client wants gold exposure for 2 years but can't lose more than 5%. What can we offer?
That reads as a participation note on Gold, 2Y, 95% capital protection. Two questions before I price it: ticket size, and should the upside be capped?
Participation noteGold2Y95% protected

02 · Price

Priced on a calibrated surface, not a flat vol.

Smile-consistent pricing off the daily-calibrated vol surface, on the issuer’s own funding curve. Change the ask, protection, cap or barrier, and the terms reprice in seconds.

About the arrowquant engine
Quotesurface: NIFTY · latest close
Tenor · Protection3Y · 100%
StrikeATM, initial fixing
Issuer fundingA1+ · 9.50%
Structuring margin2.00%
Participation110%

03 · Stress

Know what moves it before the client asks.

Two views, kept apart: the issuer’s hedge Greeks and spot × vol stress grid, and the investor’s outcome at maturity, with worst case, VaR, expected shortfall and the summary risk indicator.

Issuer MtM · spot × vol

−20%spot+20%

Investor · at maturity

Summary risk indicator

1234567

Worst case

₹0

100% protected at maturity, subject to issuer credit.

04 · Document

The KID writes itself. You still read it.

A regulation-style key information document and term sheet, with risk indicator, performance scenarios, cost breakdown and payoff profile, built from the same run that priced the note. Ready for internal review, not for skipping it.

Key information document

3Y NIFTY Participation Note

Performance scenarios

Indicative term sheet

Terms & costs

UnderlyingNIFTY 50
Tenor3Y
Protection100%
Participation110%
Total costs2.00%

05 · Review

Review with a trail behind it.

Every engine call is logged against the access key that made it: which tool ran, when, whether it succeeded, and the conversation it came from. Product heads review from there. Revoke a key and it stops working at once.

Audit traildesk: private-wealth
09:12:04span_quote · participation200
09:12:31span_compute_risk200
09:12:58span_generate_kid_pdf200
09:13:02trace tr_…9c1 · chat turn linkedlinked
10:02:57key revoked by adminrevoked

How it works

Work where you already work. One engine underneath.

Chat, your AI assistant, the API and the Structures shelf all call arrowquant. A quote in chat matches the shelf and the KID to the paisa.

OpenArrow chat

Conversational structuring

Your AI assistant

61 tools over MCP

API

For platforms and agents

Structures shelf

17 ready templates

Engine

arrowquant

Vectorised end-to-end, deterministic
Daily-calibrated vol surfaces
Forward, discount and funding curves
Monte Carlo and closed form

Quote

Terms, fair value, costs

Risk

Greeks, stress grid, risk indicator

KID & term sheet

PDF, review-ready

Bespoke payoffs

Payoff graphs priced on the same paths

Getting started

Your first 7 days.

Day 1

Account, keys, connected

Keys are provisioned from the dashboard and your AI assistant links in minutes. Nothing to install on your side.

Day 2

Price the shelf

Work through the 17 templates and the structures your desk already sells, on today’s surface.

Days 3–6

Pilot on your book

Run live briefs through your own review process, side by side with how you do it today.

Day 7

Decide on evidence

The usage dashboard shows what was priced and documented, and by whom. Then you decide.

Education

Teach structuring the way desks do it.

Colleges and universities can license OpenArrow for classroom teaching, labs and research, starting this academic year.

Students design, price and document structured products by talking to an AI agent, working with calibrated market data and the same pricing engine that powers OpenArrow's production platform.

Classroom

Build structured products such as reverse convertibles in conversation, then examine the term sheet the engine produces. Every number is engine-computed, not LLM-generated.

Labs

Explore volatility surfaces, Greeks, scenario analysis and key information documents through hands-on exercises on daily-calibrated data. Change assumptions, rebuild, reproduce.

Research

Programmatic access to the pricing engine and historical volatility surfaces for derivatives, fintech and AI research, with usage tracked by seat for labs and research groups.

Academic licence per programme, per yearFor financial engineering, derivatives, fintech and AI programmes
Academic licensing

Controls

The model talks. The engine computes.

Language models are good at understanding a brief and bad at arithmetic. So they never produce a number.

Deterministic

Same inputs, same numbers

Reproducible pricing on a recorded surface snapshot.

Engine-computed

No generated figures

Every figure in a quote, KID or term sheet comes from arrowquant.

Governed

Per-key audit and revocation

Usage metered and logged per seat, revocable at once.

Enterprise

Bring your own engine. Bring your own data.

Run your validated models and keep proprietary data in-house, behind OpenArrow's workflow and agent layer. Scoped per deployment on enterprise plans.

Pricing

Tailored to your team.

Every desk is different. Tell us about your workflow and we'll put together the right plan, free evaluation included.

Questions desks ask first.

All FAQs
Does the language model produce any of the numbers?+

No. The model interprets the brief and calls the engine. Every price, Greek, scenario and cost is returned by arrowquant and passed through unchanged.

Which underlyings and payoffs are supported?+

NIFTY and gold today, with daily-calibrated vol surfaces. Ten payoff families cover capital-protected and yield-enhancement notes, and the Structures shelf ships 17 ready templates, including bespoke Indian notes described as payoff graphs. Additional underlyings are added on client demand.

Can we use our own pricing models or data?+

Yes, on an enterprise plan. Running your validated models and keeping proprietary data in-house behind the workflow layer is scoped as part of a dedicated deployment. Tell us what you run today and we will map it.

Is the KID ready to hand to a client?+

It is ready for internal review. The document carries the risk indicator, performance scenarios and cost breakdown from the same engine run, but final terms and client-facing documents always come from the issuing institution.

Which AI assistants can connect?+

Any workspace that supports the Model Context Protocol, Claude, Cursor and VS Code among them, through a one-time connection that takes a few minutes. OpenArrow chat needs no setup at all.

Put one product through the workflow.

Bring a real brief from your desk. We'll run it end to end in 60 seconds, brief to review-ready KID.