PRE-SEED · USD 300K · OWN MODEL IN PRODUCTION
NapsixB2B
Own model
We shipped Mato v1.0: our first proprietary model, served on our own H100s. The free floor stops being a cost and becomes margin — and the platform already running in production distributes it.
The agentic OS with its own inference: 1000+ apps, APIs and MCPs, and 300K+ businesses already accessible via partners.
Per-token AI broke the budgets. Microsoft and Uber had to shut Claude down.
May 2026: the experiment ended. Microsoft cancelled ~100,000 Claude Code licenses and Uber burned its entire 2026 AI budget in four months — not because AI doesn't work, but because they routed EVERYTHING through the most expensive model. The uncomfortable truth: for more than 90% of office work you don't need the trendy model; an efficient open-source model solves it just as well. That's exactly what NapsixB2B delivers: it routes each task to the right-sized model, with a free-model floor always on to stay available 24/7.
- 100K
Claude licenses Microsoft shut off
It cancelled ~100,000 Claude Code licenses (deadline Jun 30) after burning its annual AI budget in months — despite having invested USD 5B in Anthropic.
The Verge · May 2026 - 4 months
How long it took Uber to burn its 2026 AI budget
The CTO warned: “the budget I thought I'd need is already blown.” Engineers spending USD 500–2,000/mo on tokens — everything routed to the most expensive model, even for tasks that didn't need it.
AI Magazine · 2026 - 85%
Companies that missed their AI cost forecast (>10%)
And NVIDIA confirms it from the inside: “compute already costs more than the employees.” 84% had to cut other budgets to pay for AI.
Mavvrik · Axios · 2026 - $725B
Big-4 AI infra spend 2026
Investment is at record highs, but spending it by routing everything to the most expensive model is unsustainable. What the market needs is the opposite: the right model per task. That's Napsix.
2026
The timing: AI cost already split those who rent from those who own.
Microsoft and Uber shut Claude down over cost. We shipped Mato v1.0: our own inference for 90% of office work — at near-zero marginal cost on our H100s — with frontier models (Claude, GPT, Gemini) only where they're truly needed. We serve the free floor ourselves: efficiency by design, not brute force.
AI arrived. The budget exploded.
The token trap
Per-token pricing is unpredictable: agents read files, execute, retry — and the meter runs when nobody's looking. Microsoft and Uber found out at fiscal year-end. 85% of companies missed their AI cost forecast by more than 10%.
One vendor = one point of failure
If Claude goes down, raises prices or rate-limits you, your agents stop. Anthropic already stripped bundled tokens from enterprise plans. Locking into a single model is a business-continuity risk, not just a cost one.
Fragmentation and stuck teams
ChatGPT + Claude + Gemini + Copilot + Cursor ≈ USD 99/user/month: a team of 10 pays USD 11,880/year, 5 invoices in 3 currencies. And AI only reached engineering — in sales, ops and support fewer than 20% use it.
Distribution doesn't scale
SaaS-AI vendors burn USD 800–1,500 CAC per SMB. Companies don't respond to outbound. The winner needs a different distribution model — not more reps, but pre-built networks with thousands of businesses ready to activate.
“AI arrived.
The budget exploded.”
Own model + multi-model. The free floor turns from cost into margin.
NapsixB2B is the agentic OS with its own inference: Mato v1.0 serves 90%+ of office work from our H100s (verified tool-calling, 256k context, $0 per token on every plan) and routes to frontier models only where needed. 1000+ apps, APIs and MCPs connect your team. From the founder to the corporation — and it distributes through networks, not seat by seat.
LIVEOwn model + multi-model, free floor 24/7
Mato v1.0 (ours, on our own H100s) serves the free floor at near-zero marginal cost; automatic routing adds Claude, GPT, Gemini and open models by cost and task. If a provider goes down or gets expensive, your agent keeps working: the meter never leaves you without service.
LIVEThe right model for each task
For 90%+ of office work you don't need the trendy model: Mato v1.0 solves it with native tool-calling over Napsix's tools, at a fraction of the cost and latency. The frontier ones (Claude, GPT) are saved for what truly needs them. Efficiency by design, not brute force.
LIVETeams + 1000+ apps + ecosystems
Rooms where people and agents work together, with shared memory. 1000+ apps with managed OAuth open access to huge ecosystems: Tiendanube (180,000+ brands), Shopify, MercadoLibre and more. RBAC, audit log and enterprise compliance.
LIVEAPIs + MCPs = open ecosystem
OpenAI-compatible gateway + public MCP servers: any app or AI client can operate on Napsix. Network-ready architecture (white-label, multi-org, SSO): one platform deal = thousands of businesses onboarded. The best ecosystem for companies of any size.
XIA Agents — the heart of the stack
Chat + Agents + Automations + CRM + Omnichannel Inbox + Calendar + Knowledge Base + Rooms for teams. Plus our own AI-native suite: Napsix Mail (self-hosted email), Drive, Forms, Docs, Concierge, Dashboards (AI BI), Napsix Sheets (AI spreadsheet), XIA Crawl (web search) and Mato v1.0 (proprietary model) — with Design on the way. 1000+ apps with managed OAuth, APIs and MCPs. Multi-model over AWS Bedrock + Vercel AI Gateway with a free 24/7 floor. From the founder to the corporation.
Live. Real production. An own model + an AI-native suite.
Mato v1.0 is our first proprietary model: it runs on our own H100s, with verified tool-calling over Napsix's tools and 256k context, at $0 per token on every plan. Around it, the full suite in production with real users: Napsix Mail, Drive, Forms, Docs, Concierge, Dashboards, Sheets and XIA Crawl, with Design on the way. Built in-house on Next.js 16 + Postgres + multi-model gateway + proprietary integration platform.
Mato v1.0 — proprietary model
Served on our own H100s: native tool-calling over Napsix's tools, 256k context, ~30 tok/s and $0 per token on every plan, with automatic fallback. The free floor stops being a cost and becomes margin.
A suite of AI-native products
Self-hosted Napsix Mail, Drive, Forms (AI input), Docs, embeddable Concierge, Dashboards (AI BI), Napsix Sheets (AI spreadsheet, single-editor) and XIA Crawl (web search) — every product AI-native. Design on the way.
Visual flow + 1000+ apps
Automations, B2B templates and a dynamic integrations node. Email, WhatsApp, schedule and webhook triggers.
CRM + Inbox + Calendar + Rooms
Kanban pipeline, omnichannel inbox with WhatsApp, calendar, knowledge base and rooms for teams. All native, all in one place.
XIA Agents — Agentic Stack
Chat + Agents + Automations + CRM + Inbox + Calendar + KB + Rooms. Plus the AI-native suite: Mail, Storage, Forms, Documents, Concierge, Dashboards, Sheets and XIA Crawl. 1000+ apps with managed OAuth, APIs and MCPs. Own inference: Mato v1.0 on Daytona's H100s (Startup Grid Program) with Alibaba Cloud's Qwen family as the model's base, plus multi-model routing and a free 24/7 floor.
Mato v1.0 is our first proprietary model: it runs on our own H100s, with verified tool-calling over Napsix's tools and 256k context, at $0 per token on every plan. Around it, the full suite in production with real users: Napsix Mail, Drive, Forms, Docs, Concierge, Dashboards, Sheets and XIA Crawl, with Design on the way. Built in-house on Next.js 16 + Postgres + multi-model gateway + proprietary integration platform.
The skills are built in
Mato needs no plugins: the chat commands (/web, /doc, /sheet, /crm…) are part of the model. Every new skill ships to all customers at the same marginal cost — the moat isn't the feature, it's the cost of serving it.
/image/edit/video/reel/doc/sheet/dashboard/site
One integration → huge ecosystems
Through 1000+ apps with managed OAuth we reach massive ecosystems: Tiendanube (180,000+ brands), Shopify, MercadoLibre, and the whole productivity stack. APIs and MCPs open the rest.
I have prepared the next steps.
- Gmail
- Slack
- HubSpot
- Google Sheets
- Stripe
- Shopify
- Tienda Nube
- Notion
- Trello
Live product at napsix.com
Network TAM: 300K+ businesses already in our pipeline.
Traditional bottom-up TAM says USD 6B. Our operable TAM today is 300,000+ businesses already accessible via signed agreements — including the Tiendanube ecosystem. And with 1000+ apps, APIs and MCPs we reach even larger ecosystems (Shopify, MercadoLibre and more). Not a market to conquer — a network to activate.
| Partner | Context | Businesses | Status |
|---|---|---|---|
| Tiendanube App Store | Official app (ID 32657) · 180,000+ brands across LATAM · via 1000+ apps also Shopify, MercadoLibre and more ecosystems | 180,000+ | Official app |
| Negozona (Franchises) | LATAM franchise network · 165 active franchises · 20K operating locations | 80,000 | Active partner |
| ASEA (Fintech AR) | Argentine Fintech Association · Germán is founding partner | 40,000 | Founder partner |
| AWS Marketplace | Global listing · Co-sell motion with AWS · Enterprise tickets | Global | In approval |
| Pygma Portfolio Perk | Active perk for every Pygma portfolio startup · Founders-friendly | Portfolio | Active perk |
- 300,000+
- TOTAL ACCESSIBLE
- Revenue USD 5.7K
- Activable from day one
- $200M ARR
Cursor
Code copilot · PLG-first
- PLG → ent
n8n
Automation · open-source → enterprise
- 40% share
Anthropic
Enterprise API · 18 months
Who signs and who uses it
Buyer: owner or GM of an SMB with 10 to 500 employees (decides and pays). Daily users: operations, sales and support teams — the ones already using AI outside the platform.
Each partner already has a distributed network of thousands of businesses trusting their brand. We activate their base with a single deal — no SDRs, no outbound, no per-seat selling. Tiendanube opens 180,000+ brands; the 1000+ apps open the rest.
Four engines. One network distribution.
- Engine 1 · Anchor
Network Platform Fee
Monthly fee per activated network. The anchor revenue stream — one deal = thousands of users. Priced by network size, not by individual user.
- Network StarterUSD 2.5K/mo · up to 500 locations
- Network GrowthUSD 8K/mo · 500–5K locations
- Network EnterpriseUSD 20K+/mo · 5K+ locations
- Engine 2
Multi-model token markup
Margin on AWS Bedrock + multi-model gateway. ~75% gross margin. We absorb the price volatility across providers and the customer gets predictability + a free floor. Linear scale with actual usage.
- ✓Transparent markup on tokens
- ✓~75% gross margin · we absorb the volatility
- ✓Free 24/7 floor included — the customer never loses service
- Engine 3
Vertical Packs
Industry add-ons: USD 500–2K/mo on top of the platform fee. Sector customization with pre-built knowledge base, agents and flows.
- →Retail · E-commerce · Franchises
- →Fintech · Manufacturing
- →Pro services · Healthcare
- Engine 4
Marketplace + Ecosystem (APIs/MCPs)
Co-sell via AWS Marketplace + free distribution through APIs and MCP servers. The gateway and MCPs turn Napsix into infrastructure: any app builds on us. Enterprise ticket USD 50K+/year.
- →Co-sell with AWS PSM · ticket USD 50K+/year
- →OpenAI-compat gateway + public MCPs
- →Ecosystem of apps building on Napsix
Philosophy: Engine 1 anchors the economics. The rest compound.
Engine 1 (Network Fee) generates predictable revenue and validates demand · Engine 2 (token markup) compounds with real usage · Engine 3 (verticals) deepens margin · Engine 4 (Marketplace + APIs/MCPs) opens the enterprise pipeline and the developer ecosystem.
The economics that break the market.
Traditional SaaS-AI burns CAC on outbound to businesses that don't respond. Our network model flips the math: one platform deal activates thousands of businesses with one close.
Traditional B2B SaaS
- CAC per businessUSD 800–1,500
- Sales cycle60–90 days
- Businesses per deal1–10
- LTV / CAC~3x
Napsix · Network Model
- CAC per network dealUSD 15K
- Businesses per deal20,000 avg
- Effective CAC per businessUSD 0.75
- LTV per business (3yr)USD 1,800
The Negozona deal alone unlocks 80,000 businesses with zero SDRs, and Tiendanube adds 180,000+ brands as an official app. Distribution is the moat.
Everyone sells intelligence. Nobody is multi-model, with a free floor and for teams.
| Distribution model | ChatGPT / Claude | NAPSIX ★ |
|---|---|---|
| ChatGPT / Claude | Self-serve individual | Network-led (B2B2B2C) |
| GitHub Copilot | Per-seat (Microsoft) | Network-led (B2B2B2C) |
| Cliengo / Aivo | Per-account sales | Network-led (B2B2B2C) |
Seats
ChatGPTClaudeSingle-model
Seats
Copilotn8nMulti-model
Teams
CliengoAivoSingle-model
Teams + Ecosystem
NAPSIX ★Multi-model + free
Axes: Single-model vs Multi-model + free × Seats vs Teams + Ecosystem
Fintech veterans, building for this moment.
Founder-Market Fit
No pivots: the same fintech and risk expertise, evolving from Global Group through CobroDirecto, Olpays and Ohana into Napsix's agentic multi-tenant OS. The team built XIA Agents in production, in-house, no outsourcing. Networks already trust the founders — Tiendanube, ASEA, BCRA, Negozona, Pygma.
We didn't pivot. We evolved.
- 2009Global Group · risk & fintech for banks
- 2009–21CobroDirecto → Olpays → Ohana
- 2021Napsix brand acquired
- 2025XIA Agents · agentic multi-tenant OS
- 2026NapsixB2B · multi-model + 300K+ network
Backed by the AI giants. USD 125K in non-dilutive grants and credits.
Anthropic and AWS — the two players dominating the agentic stack — chose Napsix before the round, Tiendanube homologated us as an official App Store app, and Daytona brought us into its Startup Grid Program. External validation + runway extension + strategic alignment with the infrastructure we already run on — the same one serving Mato v1.0, our first proprietary model.
- USD 15,000
Anthropic
Anthropic GrantRecognition of our work building production-grade agentic AI on Claude for companies in every industry. Direct credit on one of the models we route to.
Awarded - USD 10,000
AWS Activate
AWS Activate CreditsApproved on November 4, 2025. Credits for AWS Bedrock + production infrastructure. Co-sell motion enabled with the AWS Startups team.
Approved · 11/04/2025 - USD 100,000
Daytona
Startup Grid ProgramNVIDIA H100 infrastructure credits to serve Mato v1.0, our first proprietary model. Recognizes the agentic platform already running in production with real customers.
Awarded
What this unlocks
- USD 125K in grants and credits extending runway with zero dilution
- Official Tiendanube app (ID 32657) — 180,000+ brands accessible
- Strategic alignment with the leading labs (Claude 40% API share · AWS Bedrock enterprise leader)
- First proprietary model (Mato v1.0) served on the program's infrastructure
- External validation signal for follow-on capital
Neither an AI lab nor a 180,000+ brand marketplace homologates projects without product-market evidence. Anthropic, AWS, Tiendanube and Daytona bet on us before the round — investors enter on a validated track.
USD 300K to activate the network. 10 network deals · 300K+ businesses accessible in 12 months.
Use of funds
Milestones · 12 months
Month 3 · Activation
Mato v1.0 in GA (vision + 1M context). 3 network deals signed (Negozona pilot live). 2,000 active users (people using the platform). MRR USD 8K.
ActivationMonth 6 · Traction
Second generation of the model under evaluation. 6 network deals signed. 8,000 active users. MRR USD 35K. First co-sell via AWS Marketplace.
TractionMonth 9 · Scale
8 network deals. 18,000 active users (people inside the activated accounts). MRR USD 75K. NPS > 50.
ScaleMonth 12 · PMF
10 network deals · 300K+ businesses accessible. 12,000 active businesses in the network (accounts, not seats). MRR USD 120K · ARR run-rate USD 1.4M. Series Seed pipeline.
PMF
Two different metrics: active businesses = accounts activated through the network (the buyer who signs); active users = people using the platform inside those accounts. The network activates by account, not by seat.
Applicable RIMI benefit
As an Argentine tech micro-company, Napsix qualifies for RIMI (min. USD 150K). Accelerated income tax amortization + early VAT refund within 3 months. Reduces the real cost of investment for the fund and gives us a regional fiscal lever.
The agentic operating system for companies in every industry.
- 2026 H2
Mato v1.0 in GA (vision + 1M context) serving the free floor, multi-model + efficiency routing, Tiendanube ecosystem active. First paying enterprise customer via AWS Marketplace.
- 2027
Second generation of the model + vertical models trained on proprietary data. Global expansion via AWS Marketplace + e-commerce ecosystems (Tiendanube, Shopify, MercadoLibre). 1,000 active networks, vertical packs in production, APIs/MCPs as a distribution channel.
- 2028
Every industry — from the founder to the corporation. Vertical models per industry and own inference as structural margin. Stack open for vertical partners. Enterprise mid-market consolidated via co-sell with AWS and Anthropic.
- 2029+
Global standard for enterprise agentic AI: the place where companies operate with AI without cost risk and with their own models. Data products + intelligence layer shared across verticals.
“Every conversation that flows through Napsix makes the customer's team more productive — with no risk of the bill exploding. With our own model, a free floor and frontier where it matters. That's real AI adoption.”
USD 300K to activate the network.
10 network deals · 300K+ businesses accessible in 12 months.
Your next step
From evaluation to a concrete integration.
Start with an account or talk to us about the scope of your team and application.
