Kryptos 2.0: built for AI agents, payment institutions, and billions of transactions

Kryptos 2.0 is a complete rebuild of the platform underneath Kryptos, built for where finance is heading: AI agents working directly on financial data, and institutions that move crypto at scale, from treasury teams and funds to payment institutions handling millions of transactions and beyond. Here is what changed, and everything new in the enterprise product.

Why we rebuilt the platform

A corporate treasury does not use a crypto tax tool the way an investor does. There are entities and subsidiaries instead of one wallet list. Several people work the same ledger with different levels of access. Reports get filed with a regulator and then have to stay exactly as filed while the team keeps working in the same data. And an auditor will ask who changed a number in March and expect a real answer.

Kryptos began as a tax product for individual investors, and the original platform did that job well: connect a few wallets, bring in a few thousand transactions, file once a year. It could not stretch to the job above. A single treasury account can hold more activity than hundreds of those early accounts put together.

So instead of patching it further, we rebuilt it. Kryptos 2.0 is a new foundation underneath the whole product, made for accounts that are large, shared between people, and audited.

We rebuilt it for where finance is going, not only for where our accounts are today. Two shifts shaped the design. The first is scale: payment institutions, exchanges and large funds generate transaction volumes that dwarf a normal treasury, running into the millions and beyond, and the system has to keep answering questions instantly at that size. The second is automation: financial work is moving toward AI agents that read data, prepare reports and reconcile activity on their own. A platform built for that era has to expose everything it can do through clean, permissioned interfaces, rather than leaving it behind a screen a person has to sign into. Kryptos 2.0 is built for both.

This was a total rewrite of the platform, not a refactor. We rebuilt it from the database up, because the limits we kept running into were in the shape of the old system rather than in how we had tuned it. A rewrite is expensive, and we would not have picked one if patching could have got us there. The difference now is that a bigger account makes the system do more work, instead of running it out of room. Akash Srivastava, Chief Technology Officer, Kryptos

What the rebuild changed

Six things are different at the foundation, and they are the reason the features further down could be built at all.

A new store underneath everything

The database we started on was designed for small, self-contained accounts. It could only answer questions we had set it up for in advance, which is why filtering came in fixed combinations and totals took a while to appear. An enterprise relational database now sits under the whole product. It can answer questions we did not anticipate, which is what makes free-form filtering, comparisons across wallets and immediate totals possible at all.

Room for far more activity

Large accounts used to slow down and eventually stall somewhere past a few hundred thousand transactions. Kryptos 2.0 is built for millions of transactions in a single account. Capacity now grows with the work instead of running into a fixed limit, so a busy month no longer means a slow product.

Only recalculate what changed

Correcting one transaction from two years ago used to mean recalculating your whole history, so a small fix could mean a long wait. Kryptos now saves its work as it goes and picks up from just before whatever you changed. A correction to an old transaction costs a fraction of the time it used to.

Built for large accounts

Screens that had to be reloaded to tell you anything had changed are gone, and so is filtering limited to a handful of fixed combinations. Progress updates live while work is running, filters combine freely, and the layout is designed for working through thousands of rows rather than reviewing a few dozen.

Errors you can actually see

A failed sync used to look the same as nothing happening, and you would find out from a number that looked wrong later on. Now every operation records what it attempted, how far it got and what went wrong. Failures surface on the integration itself and in the audit trail, with enough detail for support to act on without asking you to reproduce it.

Standard interfaces, not exports. Getting data out of the old product meant downloading a file, and connecting anything else to it meant custom work on both sides. More than 140 operations are now available through a documented API, with permissions you grant per area, read or write, and revoke whenever you want. The same operations are exposed over the Model Context Protocol, the emerging standard for AI assistants, so an agent can check holdings, pull a report or reconcile a batch of transactions with your approval and inside exactly the permissions your own team has.

What is new in the enterprise product

Organizations and workspaces

Your company is one organization with its own billing details, and each entity, fund or subsidiary gets a workspace underneath it. Every workspace keeps its own country, base currency, cost basis method and timezone, so a US entity and a German one can sit side by side without either bending to the other's rules.

Roles and permissions

Seven roles cover the ways a finance team divides work, from full administrators through managers and accountants to read-only access for an auditor or an outside advisor, with separate read and write permissions for SAFT tracking. People are invited per workspace, so an entity's accountant sees that entity and nothing else. Locked periods are enforced on the server rather than hidden in the interface, which means a closed period is closed for everyone no matter how they reach it.

Live sync monitoring

Add ten chains at once and every one of them appears straight away, moving from queued to processing to a finished transaction count and value. A progress indicator stays in the header while anything is still running, so you can see what is happening without refreshing and without wondering whether it has stalled.

Sync history

Every integration keeps a record of each sync: when it ran, whether it came from an API or an uploaded file, what started it, how far it got and when it finished. Three resync options sit above it, for pulling the latest activity, starting again from the beginning, or refreshing prices only.

Lock periods

Once you have filed for a quarter or a year, lock it. Everything dated on or before that date stops affecting your accounting, so a later edit cannot invalidate a report you have already submitted after the fact. Locks release in reverse order, so nobody can reopen 2023 while 2024 and 2025 are still locked on top of it, and the history shows who locked what and when.

Audit history

A full record of what has happened in the account: what was done, to which item, by whom and when. Activity from the system and activity from people are labelled separately, so a scheduled sync does not look like something a colleague did. Filter it by type, event, person or date.

Custom asset prices

Unlisted tokens, SAFT positions and internal instruments have no market price. Set your own price per asset and Kryptos uses it instead of looking one up. The screen tells you at the moment you save that existing transactions pick up the change on the next reprocess, so there is no guessing about what has already been repriced.

Reconciliation

Reconciliation is a team activity, and Kryptos 2.0 treats it as one. Every filter works both ways: include the wallets you want, exclude the currencies you do not, add a date range, and see a count next to each value so you know the size of what you are removing before you remove it. Holdings show amount, price, value, cost basis, 24 hour change, ROI, unrealised gains and a transaction count for every asset, and you can switch to calculated balances to compare what Kryptos worked out from your transaction history against what the exchange or chain reports, which is the quickest way to spot a discrepancy worth investigating. Leave comments on individual transactions, flag the ones that need a second look, then filter the table down to just those, or to spam, ignored, DeFi, NFT, manually added or edited rows.

Also in the enterprise product

The rest of what an enterprise account includes, built on the same foundation.

• Chart of accounts and ERP sync. Map crypto activity to your general ledger and push journal entries to Xero, QuickBooks or SoftLedger. Connect your accounts once and keep them in sync.
• Rules engine. Rules that classify transactions and assign ledger accounts automatically, applied across the whole account rather than row by row.
• Reviews and reconciliation. A queue of what needs attention, with counterparty and contact matching so transfers resolve to a name instead of an address.
• Reports. Tax and treasury reports per entity and fiscal year, with the disposal detail sitting behind every number.
• Tax optimization. Loss harvesting analysis against current holdings, so a position can be reviewed before a year end rather than after it.
• SAFT tracking. Token grants and vesting schedules, with their own read and write permissions separate from the rest of the account.
• Invoicing and payouts. Invoices, recurring invoices and batch payouts, settled in crypto and recorded in the same ledger.
• Programmatic access. Scoped API keys per workspace, and OAuth for applications you connect to your account.

Everything above is live in enterprise accounts today. If you are evaluating Kryptos for a treasury, a fund or a group of entities, the quickest way to judge it is against your own data. Book a walkthrough and our team will set up a workspace per entity and connect your first integrations with you.

About the author

Akash Srivastava

Chief Technology Officer