Skip to content
Luyten
Get started(opens in a new tab)

Automated finance for 10x growth startups

Luyten’s agents run the books. A second AI checks every entry, and anything material waits for your OK.

Get started now(opens in a new tab)
The Luyten Overview for September 2026: cash, income and spend, items waiting for the CFO's OK, bank lines to decide before the close, the KPIs, proposals by status, close progress, agent status and customer credit. Playing, it shows the CFO starting the close, the agents working, and the CFO confirming a proposal.

Built on real data, made for lean finance teams

Data sets

Bureau, identity and public-footprint data behind every credit check

  • LinkedIn

Company types

The finance teams Luyten is built for

  • SaaS companies
  • Selling to small businesses
  • Groups of companies
  • Multi-currency books
  • Solo controllers
  • CFOs with a small team

Technologies

The models and infrastructure under the hood

  • xAI
  • AI SDK
  • SpacetimeDB
  • Svelte
  • TypeScript
  • Node.js
  • Vite

Solutions

What the agents take off your plate at month end

  • Bank reconciliation
  • Revenue recognition
  • Accruals and payables
  • Consolidation and FX
  • Credit and risk
  • Board Pack

Logos are trademarks of their owners and show the data and tools Luyten uses, not endorsements.

Our automated agent workflow

How work moves between them: Bank feed to Bookkeeper; Billing to Bookkeeper; Billing to Revenue; Bureaus & web to Credit & Risk; Group books to Consolidation; Bookkeeper to Red Team; Revenue to Red Team; Credit & Risk to Red Team; Consolidation to Red Team; Red Team to Advisor; Advisor to Ledger (under $10,000); Advisor to CFO ($10,000+ or amended); CFO to Ledger.

Bank feed
Billing
Bureaus
Books
Ledger

Agent security, and the APIs behind Luyten

Agent security SpacetimeDB

Multi-layer agent security starts in the database

A worker agent proposes. The Red Team critiques, Grok first and then Gemini. The Advisor rules and writes down why. The CFO confirms anything at or above Materiality. Under all of it, SpacetimeDB reducers check every caller’s identity, so no prompt can talk its way into the ledger. The feeds can only add raw bank lines and invoices.

The gate checks who is calling. A worker agent that tries to post to the ledger is refused, whatever its prompt says; only the Advisor, below your threshold, and you, above it, get a Proposal posted.

  1. 01

    Worker agents propose

    Bookkeeper, Revenue, Credit & Risk. Each signs in as its own role. They propose, never post.

  2. 02

    Red Team critiques

    Grok attacks each Proposal first, then Gemini adds what that critique missed.

  3. 03

    Advisor rules

    Approves or rejects every Proposal and writes down the rationale.

  4. 04

    CFO confirms

    Anything at or above Materiality waits for the CFO, a seat claimed with a one-time secret.

  5. 05

    SpacetimeDB reducers check identity

    Every write checks who is calling. An agent that ignores its prompt still cannot post.

  6. 06

    Ledger

    Journal Lines, Proposals and the audit trail, streamed to the app as they change.

Credit checks

Know the people behind every customer and vendor

Before money moves, the Credit & Risk agent runs a cascade on the CRS Credit API: identity, fraud, all three credit bureaus and public records. The books derive the Risk Tier from what it finds, and the tier sets the payment terms.

01 Identity

tu_identity_verification

A real, living person who matches the name on file. A deceased or mismatched identity stops the cascade here.

02 Fraud

fraud_finder

A fraud score on the email, phone and address they gave you.

03 Three bureaus

credit_report

TransUnion, Equifax and Experian, all three, or the check is refused.

04 Public records

pitchpoint_blj · eviction · criminal

Bankruptcies, liens, judgments, evictions and criminal records.

The books set the tier

The database derives it from the recorded facts, so the agent cannot claim a friendlier one.

  • A 45 days
  • B 30 days
  • C 15 days
  • D No credit
  • BLOCK No credit

The result becomes a Proposal, so the Advisor rules on it and the CFO confirms it before credit is extended, reserved or blocked.

Advisor chat

Ask your Advisor how the business is doing

Luyten lives on Fetch.ai’s Agentverse, so you can chat with it in ASI:One. It asks four specialist agents at once and answers with one interactive card, built from the live books.

Luyten
ASI:One · Agent Chat Protocol 0.3.0

How are we doing?

Asking four specialists at once: BookkeeperRevenueCredit & RiskConsolidation

Financial health check

  • Cash and the bank Cash per company in its own currency, and unexplained bank lines
  • Revenue and receivables Revenue recognized this month, what customers owe, what is overdue
  • Credit risk Customers by risk tier, the bad-debt reserve, vendors blocked before payment
  • The group and FX The month’s exchange rates, revaluation and elimination entries
What needs the CFO’s OK? Why? Start the close

It reads the books and explains them. It never confirms or posts anything: the CFO does that in the Luyten app.

Public footprint

See what anyone can see, before you extend credit

For each customer and vendor, Bright Data collects what a logged-out visitor can see: a Yelp listing through Web Unlocker and a LinkedIn company page through the Web Scraper API.

Signals for a person, not a decision
What a logged-out visitor can see

  • Bansho Ramen · Customer

    Yelp listing · Web Unlocker

    Open on Yelp with 4.1 stars from 196 reviews, and the owner has claimed the page.

    7 more reviews since the last run

  • Stripe · Vendor

    LinkedIn company page · Web Scraper API

    Lists 16,694 employees on LinkedIn and has about 1.8 million followers. Founded in 2010.

Every run is stored, so you see what changed. Credit & Risk cites these in its reasons, never in a tier by themselves, and no owner, reviewer or poster names are kept.

Built for your 10x

A startup that grows 10x outgrows its finance stack on the way. Luyten grows with you from a $10M seed to a $1B growth round, and the gate never loosens as the volume climbs.

Worked example: Quillmesh

Our sample company sells a document-automation API. Series A at $50M post-money, about $4.8M ARR growing about 3x a year, a US parent and a UK subsidiary. It has just entered stage 2.

  1. Stage 1 Available now

    $10M to $50M

    Seed to Series A

    What changes for finance

    One or two entities, a few hundred customers, and the founder still approves spend.

    What Luyten does

    Agents propose, the Advisor rules, and the CFO confirms anything material. Close in days, not weeks. Cap table and round modeling.

  2. Stage 2 Available now

    $50M to $250M

    Series A to Series B

    What changes for finance

    A second country and currency, usage pricing, enterprise contracts, and a first controller.

    What Luyten does

    Multi-currency consolidation, revaluation and eliminations as Proposals. Revenue schedules rebuilt automatically. Credit checks on customers and vendors. Custom Charters that only tighten the gate.

    Quillmesh is here: a UK subsidiary in GBP, per-page usage pricing, about 15 enterprise contracts and a finance team of three.

  3. Stage 3 On the roadmap

    $250M to $1B

    Series B to growth

    What changes for finance

    Many entities, tens of thousands of transactions a month, audit and SOC 2.

    What Luyten does

    Continuous bookkeeping between closes, per-subsidiary locks, a dedicated evaluator identity, real bank and billing feeds, hosted deployment and SOC 2 controls.

The gate does not loosen as volume grows: the Materiality threshold and Charters scale with the company, and every number still drills down to its journal lines.

Run your finance with us

Get started(opens in a new tab)