One Kernel. Five Domains.
A shared substrate — Petri-net execution engine, autonomous agents, immutable ledger — deployed across procurement, treasury, life sciences, telecom, and aviation maintenance. Live in weeks, not years.
Deployed verticals
Every vertical runs on the same kernel — same Petri-net engine, same agent fleet primitives, same hash-chained ledger.
Catalog requisition, vendor discovery, RFQ orchestration, three-way match, dispute arbitration, payment release. 67% auto-approval rate, ~4 minute median human time per requisition.
Buy-side state machine, liquidity locks, immutable ledger, marketplace API, dynamic discounting that flows yield to the buyer instead of the platform.
Adverse event capture, clinical trial state machines, SUSAR reporting, CAPA cycles, SOP lifecycle management — all on the same kernel as procurement.
Ingest carrier invoices, reconcile to contract terms, flag billing errors, recommend plan optimizations. Same three-way-match pattern as procurement, applied to telecom spend.
Work orders, fleet registry, component lifecycle, life-limited parts tracking, EASA Form 1 / FAA 8130-3 certificate generation. Coming Q3 2026.
The 3-way match pattern, applied everywhere money moves
Promise → delivery → bill. Every reconciliation problem is the same shape. The kernel solves it once and skins it for each industry.
Day-one ready for the procurement office of a Fortune 500
Logic is the ultimate weapon in a world of information.
Legacy AI bolts chatbots onto SAP forms. We deleted the forms. The kernel is a database as an execution engine — PostgreSQL constraint layer with PL/pgSQL triggers, a fleet of 9 autonomous agents, an immutable hash-chained audit trail. Adding a new vertical is a route group + a state-machine INSERT, not a rewrite.
Architecture deep dive →From request to ledger entry, autonomously
Telegram requisition → PO → goods receipt → invoice → payment.
- User: /req 4 boxes A4 paper
- requisition-agent classifies + matches catalog
- po-agent issues PO
- idoc-architect emits ORDERS05
- invoice-matcher 3-way matches on receipt
- Treasury releases payment
No catalog match — agent discovers vendors, runs RFQ, awards.
- User: /req 5000 custom-printed totes
- requisition-agent: no catalog match
- rfq-agent shortlists 5 vendors
- rfq sent, bids received
- human approves shortlist
- po-agent awards & issues PO
Buyer offers early payment; supplier accepts at discount.
- Treasury detects float
- pricing-agent computes 8.4% APR offer
- supplier-agent posts to portal
- supplier accepts
- Treasury settles + books discount
Four ways to wire AI into procurement. One survives audit.
The orchestration layer is where workflow, state, and proof live. Most stacks treat it as an afterthought — a prompt loop, a doctype form, or a database trigger. The kernel makes it the spine.
vs SAP Ariba · vs Coupa · vs ERPNext
We invert the SaaS revenue model — discount yield flows to the buyer, not the platform. Open-source ERP wins on license cost; the kernel wins on autonomy and proof.
| Feature | SAP Ariba | Coupa | ERPNext | AIPROC |
|---|---|---|---|---|
| Auto-triggering discounting | ✗ | manual | ✗ | ✓ autonomous |
| Supplier enrollment | platform-managed | platform-managed | self-serve (manual) | self-serve (agentic) |
| Transaction fees | 0.1–0.35% | 0.1–0.25% | €0 | €0 |
| GAAP ledger booking | external | external | native (manual JE) | native (autonomous) |
| Hash-chained audit trail | ✗ | ✗ | ✗ | ✓ SHA-256 Merkle |
| Deterministic replay | ✗ | ✗ | ✗ | ✓ conformance harness |
| Autonomous agent fleet | ✗ | pilot | ✗ | ✓ 14 agents (P2P) |
| Human touchpoints / req | 7+ | 5+ | 5+ | ~1 |
| Discount yield destination | platform | platform | n/a | buyer |
| Marginal cost per order | €2–6 | €1.50–4 | €1.20–3 (FTE-bound) | €0.05 |
| License cost | €180k+/yr | €140k+/yr | €0 (GPL-3.0) | OEM / SaaS / per-txn |
| Implementation time | 12–24 mo | 6–12 mo | 4–12 wk | 4–6 wk |
If you go open-source, which one?
Four serious open-source ERPs that buyers consider against the kernel. None ships a hash-chained ledger or a real autonomous fleet — that's the gap the kernel was built to close.
| Dimension | ERPNext Frappe · Python | Odoo community + paid | Tryton Python · GPL-3 | Dolibarr PHP · SMB-focused |
|---|---|---|---|---|
| License Source license. Determines what you can sell/redistribute. | ✓GPL-3.0 (open core) | ⚠LGPL-3 + Enterprise | ✓GPL-3.0 | ✓GPL-3.0 |
| Stack What you maintain in production. | ✓Frappe (Python/JS) | ✓Python + OWL/JS | ✓Python (PostgreSQL) | ⚠PHP + JS |
| Procurement depth RFQ + PO + supplier scorecard + blanket + tender. | ✓RFQ, PO, scorecard | ✓RFQ, PO, tenders, blanket | ⚠PO + supplier products | ⚠PO + supplier prices |
| Multi-vertical Same kernel re-skinned for life sciences, MRO, telecom. | ⚠Healthcare module | ⚠Industry editions | ⚠Modular per vertical | ✕SMB-general only |
| Hash-chained ledger Tamper-evident audit primitive at the database layer. | ✕— | ✕— | ✕— | ✕— |
| Autonomous agents LLM-driven workers that advance state without humans. | ✕— | ⚠AI bolt-on (paid) | ✕— | ✕— |
| Pricing Total cost when self-hosted at 10k req/mo. | ✓€0 (Frappe Cloud €1.3k/yr) | ⚠€0 community / €31/user/mo | ✓€0 | ✓€0 |
| Community GitHub stars · proxy for ecosystem depth. | ✓33.9k stars · 11.2k forks | ✓45k+ stars (largest) | ⚠self-hosted Heptapod | ✓6.6k stars |
The kernel imports the best of each — vendor pricelists + supplier scorecards + blanket orders + quality inspections — and runs them through the agent fleet. See Modules.
Modules ship enabled. You turn off what you don't want.
Includes ERPNext-derived primitives — Supplier Scorecards, Blanket Orders, Quality Inspection — wired through the agent fleet and hash-chained ledger.
Most teams cut platform cost by ~67%.
Switch to ERPNext below to see the open-source comparison — license drops to near-zero but FTE load stays. The kernel removes the FTE.
Six phases shipped. MRO is next.
15-slide investor deck
Architecture, market thesis, revenue model, growth plan. Free · No form required.
Ready to deploy the kernel in your domain?
Sovereign deployment, scoped engagements, invitation-only pilot cohort for Q3 2026. Tell us what you're trying to fix.