Know what every dollar your agents spend or earn was for, and get it into your books.
Luca connects an agent’s declared intent with the payment outcome. Each payment becomes one signed record, tied to the mandate that authorized it, on any payment rail, and posted to your ERP with drill-back to the underlying evidence.
One payment, four views. Each party sees only the fields the record's owner has shared with it. Illustrative example with fictional companies.
Your agent records what it intends to pay, the rail reports what settled, and Luca joins the two. It gives every payment its accounting meaning and seals it, then posts to your ERP and shares the record with the people who need it.
Connectors copy a transaction into a ledger. Agent payments need more: evidence of who allowed the agent to spend, agreement with the other side, one ledger that no payment provider owns, and records that hold up at millions of payments a month.
A research agent at a US company pays a data provider in Ireland 0.40 USDC over x402 for one dataset query. Nobody on either side touches it.
Luca maps most payments on its own. The rest are grouped by cause, such as an unknown wallet, so one decision clears a whole group and becomes a rule for next time. A sealed record is never changed: a fix is a new sealed record, and posted entries are reversed and re-posted.
Auditors sample a few dozen invoices because checking thousands by hand costs too much. With every agent payment signed and tied to its mandate, your auditor can test all of them, through time-limited, read-only access you grant.
XBRL gave companies, auditors and regulators one language for financial reporting through the US GAAP and IFRS taxonomies. Luca applies those taxonomies to each agent payment as it happens, so every record that reaches your ERP, your auditor or the other side's agent already carries its reporting meaning.
Before paying, the agent calls record_intent with who it works for, the mandate, the purpose and the amount. Luca checks that against the mandate and budget and flags anything outside them. The payment goes ahead either way, because Luca never sits in the payment path. After paying, record_settlement adds what actually settled, or the payment network's own feed fills it in. Luca joins the two halves, seals the record and posts it to the ERP.
Luca runs as an MCP server, an A2A agent and a REST API. Other agents can read a label with get_label and pay for it per call over x402 or MPP. The record's owner decides who sees which fields and for how long, and every read is logged.
Each payment lands in the right legal entity and currency, translated at the moment it happened under ASC 830 and IAS 21, with intercompany payments matched for elimination. Records carry US GAAP or IFRS tags and export to XBRL GL, SAF-T, ISO 21378 and AICPA Audit Data Standards.
Nothing in production to connect until you are ready. Try Luca with sample data first, then with your own test payments.
A chartered accountant, former CFO of a Series A company, and three-time founder, he built Arnergy and AllBase Energy and participated in Techstars Detroit, powered by J.P. Morgan. His conviction that accounting will increasingly be run by machines comes from understanding both the financial reporting process and what standards such as XBRL made possible: structured financial information that software can interpret, exchange, and process.
LinkedInAI engineer and enterprise architect with 18 years across telecom network engineering, applied research and AI and data architecture. He works on agentic AI, GPU orchestration and data engineering at scale, has built AI systems at CohnReznick, and holds a US patent with more than 33 peer-reviewed publications.
LinkedInAnything else, ask us when you request access.
Luca is financial recordkeeping infrastructure for AI agents. It connects what an agent said it would pay for with what was actually paid, and turns each payment into one signed record tied to the mandate that authorized it. Luca reads payments from Stripe MPP, x402, Ramp Agent Cards, Google AP2 and Mastercard Agent Pay and posts journal entries to Rillet, NetSuite, Xero or QuickBooks. The buyer and the seller sign the same label, so both sets of books agree. Luca is independent of every payment rail and has read-only access.
Finance teams at companies whose AI agents buy or sell, and the developers who build those agents. Finance teams get records they can close the month on and hand to their auditors. Developers add two calls to the way their agent already pays.
Before paying, the agent calls record_intent with who it works for, the mandate, the purpose and the amount. After paying, it calls record_settlement, or the payment network’s own feed fills in the result. Luca joins the two, checks them against the mandate and budget, seals the record and posts it to the ERP. Luca runs as an MCP server, an A2A agent and a REST API, and it never sits in the payment path.
No. Luca reads payment records through read-only connections and writes accounting entries. It never holds funds, issues cards or controls wallets.
Each provider sees only its own payments. Luca is independent of all of them, so payments across Stripe MPP, x402, Ramp, AP2 and Mastercard land in one ledger, booked the same way, and no payment provider controls your accounting records.
Yes. Every customer starts in a private sandbox with sample data, then connects test rails and an ERP sandbox. Production connections come last.
Your books work the same. Labels are matched and co-signed only when both sides take part; otherwise your record stands on its own evidence.
Only you and the parties you grant access to, such as your auditor, field by field and for a set time. Every read is logged.
A correction is a new sealed record linked to the original. Posted entries are reversed and re-posted, so the full history stays provable.
No. Luca posts summarized journal entries into Rillet, NetSuite, Xero or QuickBooks, with a link from each entry back to every payment behind it.
Companies in the United States, Canada, the United Kingdom, the European Union, Australia, China, the Middle East and Asia-Pacific, in their own currencies and reporting standards.
Tell us about your company and we will set up a private sandbox and walk your team through it. We are onboarding a small number of companies now and reply within two working days.
