Book a demo

The platform

One layer. Three surfaces. Nothing for the institution to migrate.

Odyssey packages financial products for AI — and runs the rails they execute on. Discovery, suitability, execution and reporting, inside one conversation.

01

Client side

Assistant distribution

Partner products appear inside the assistant the client already uses — live cards, portfolio-aware reasoning, one-click execution. The client signs the transaction himself; the institution's product, the institution's client, the client's assistant.

  • Live across ChatGPT, Claude, Gemini
  • Live product cards with real pricing
  • One-click execution, client-signed

02

Institution side

Client intent data

Every conversation becomes distribution data the institution has never had: who explored what, who hesitated, who never clicked. The banker gets signal, not noise — he knows who to call, and why.

  • Explored vs. funded vs. left on the table
  • Per-client mandates, editable by the banker
  • Out-of-mandate attempts surfaced as signal

03

Product side

Packaged product range

Raw primitives become finished products — cash yield accounts, recurring investment plans, portfolio-backed credit, index exposure — with live pricing and suitability rules. The institution stays the manufacturer.

  • Cash yield accounts & staking live today
  • Funds, ETFs, insurance and pension wrappers next
  • Every new instrument reusable across institutions

The rails

After the click, everyone else hands off. We execute.

Enforcement at the custody layer

Eligible assets, caps and whitelists are hard-coded into the perimeter. On non-custodial infrastructure — Safe, Fireblocks, Fordefi — the mandate lives inside the policy engine itself. Out-of-mandate is not detected. It is impossible.

OdysseyRisk

Our own risk engine scores every venue, protocol and product we distribute, and gates what is allowed to reach a client's shelf. Fourteen months of live use on our own book.

Model layer

Proprietary skills, an eval suite run on every model update, a conversion corpus that compounds. Models are interchangeable; these are not.

Rail-agnostic by design

We start with programmable products because the experience is complete today — the destination is every financial product. Sequencing is deliberate: wrappers follow as the rails open.