AI agents built
for Capital Markets.

Grounded in the source. Cited to the publisher. Auditable end to end.

Request a demo Explore the agents

Built by technologists who spent decades running Wall Street systems — not by people discovering capital markets through AI.

Who we are

We designed, built, and ran these systems for over three decades

TRAC was founded by senior Wall Street technologists who spent decades building and running the trading, market-data, risk, and operations systems that capital markets depend on. We've sat in the CTO's chair and been through the compliance review — and we build accordingly.

Amit Chatterjee, Founder of TRAC

Amit Chatterjee

Founder

Wall Street technology leader building AI agents for the trading floor — grounded, governed, and desk-ready.

  • Most recently Head of Enterprise Cloud Migration & Capital-Markets Technology — U.S. Bank
  • Former Managing Director & CTO, Global Equities Technology — Citigroup
  • Former Managing Director & CIO — Weeden & Co.
  • 35+ years building trading, market-data, risk, and operations technology on Wall Street
  • Led technology diligence and integration across multiple capital-markets M&A transactions
Connect on LinkedIn Full background →

One platform. Many agents. All governed.

The information you need is public — but scattered across the SEC, FINRA, the Fed, CFTC, DTCC, and the exchanges. PRISM assembles it into grounded, cited answers, and puts governance first — every agent grounded in source documents, cited to the publisher, and logged.

Grounded & cited

Answers tied to source

Agents are built on real public filings, notices, and releases — not a general chatbot with a finance prompt. Every claim links back to the source document, with a full retrieval trail.

Assembled for you

Public, but not scattered

The data is public; the work is finding it. PRISM gathers what's spread across the SEC, FINRA, the Fed, CFTC and DTCC and turns hours of assembly into a cited paragraph.

Auditable & governed

Every call logged

Each question, retrieval, and answer is attributable and examinable — and when you connect your own systems, every call is entitlement-checked against the caller. The property that gets AI past compliance.

The Agent Catalog

Start with a live agent. Add the ones your desk needs. Each is a governed capability in one workspace — the same identity, entitlements, and audit trail across all of them.

● Live
🔎
Diligence — EDGAR

Cited Q&A and PM-ready memos over the SEC EDGAR corpus — 166,000+ filings across 500+ companies. Ask what changed in a company's risk factors and get a grounded answer with citations back to each source filing, or generate a "what changed" diligence memo across the risk and MD&A deltas.

  • Scoped, cited answers over 10-K / 10-Q / 8-K / DEF 14A filings
  • PM-ready memos built from risk-factor and MD&A change detection
  • Client-owned AWS or managed single-tenant deployment
Request a demo
● Live
🏢
Company Intelligence

A one-page, cited synopsis of any public company — assembled from its recent SEC filings and insider (Form 4) activity, with every claim traced back to source.

  • Recent filings and what stands out in the risk factors and MD&A
  • Insider activity — net buying and selling from Form 4, with notable insiders
  • Ownership (13F) and enforcement join as those sources are added
Request a demo
● Live
📋
Regulatory Change Watch

Ask what changed across the SEC, CFTC and the federal banking regulators on a topic and get a dated, cited synopsis from the Federal Register — instead of ten websites and a spreadsheet.

  • Dated synopsis of rules, proposed rules and notices, cited to each source
  • SEC, CFTC, Fed, OCC and FDIC via the public-domain Federal Register, refreshed nightly
  • FINRA and DTCC join the same workspace next
Request a demo
Preview
🏛
More agents in preview

The public record is vast and fragmented. The next agents assemble more of it into short, cited synopses — no proprietary data required.

  • Market Structure & Settlement — DTCC and exchange settlement, symbol & service changes
  • Fed / Macro Brief — FOMC, supervisory letters and key releases, distilled and cited
Request early access
Build your own

Your team composes agents.
No code.

Name an agent, write its instructions in plain English, and grant it the skills it can use. Skills are governed, public-data capabilities — SEC filings, regulatory notices, FINRA, CFTC, market-structure notices, the Fed, enforcement — each one grounded in source documents, cited to the publisher, and logged.

Example

“Weekly Regulatory Digest”

Every Monday, summarize new SEC, FINRA and CFTC notices touching equity market structure, flag anything effective in the next 30 days, and cite each source.

📋 Regulatory Notices 🛡️ FINRA Notices 🌾 CFTC 🏛️ Market Structure
  • Composable — the platform is the container; each agent is instructions plus a set of governed skills.
  • Cited by default — every answer links back to the source document, so it can be trusted and checked.
  • Governed at the skill — every call is grounded in source and logged, on every request.
  • Grows with you — new public sources drop into the same catalog, and you can connect your own systems later under the same controls.

The TRAC approach

The same principles run through everything we build. They're what make AI trustworthy enough to put into production at a financial institution.

Human-in-the-loop

Agents run the repetitive work; people keep judgment and accountability. Nothing high-impact happens without review.

Vertical & grounded

Built around real capital-markets sources — filings, regulatory notices, market-structure changes, the Fed — not a general assistant.

Cited & auditable

Every answer links to its source document and every call is logged — attributable end to end. Connect your own data later, and each call is entitlement-checked against the caller.

Client-owned & secure

Deploy in your own AWS or a dedicated single-tenant instance. Your data stays isolated; we support the machine, not the investments.

How we work: design partnership, not a vendor pitch

We build with a small number of institutions rather than selling to many. A staged engagement where each phase produces something real and an explicit decision to continue.

1

Discovery

Align on the workflows, systems, and governance requirements that matter most in your operation.

Weeks 1–4
2

Build

Stand up a working agent against one real workflow, in an environment that fits your security posture.

Weeks 5–12
3

Pilot

Run it with real users and data under supervision; measure the impact and harden the controls.

Quarter 2
4

Scale

Extend to additional agents, systems, and teams under a roadmap shaped together.

Beyond

Start a conversation

Whether you want to see PRISM live, scope an agent for your desk, or explore a design partnership — tell us what you're working on.