# NYO by Niyolab — One Governed Gateway Powering Filipino Civic and MSME AI Agents > How NYO (llm.nyolab.ai), the governed AI gateway built by Niyolab, lets Filipino builders run civic assistants, language-community agents and a one-person resort on frontier open models through one API key and one endpoint. **Source page:** https://philippineaireport.com/case-studies/nyolab **Canonical partner source:** https://llm.nyolab.ai/case-studies Category: AI Infrastructure / Sovereign AI · 2026 · 4 production agents · Open-weight models ## Problem Filipino builders of civic and small-business assistants need current open-weight models with routing, guardrails and cost accounting, without standing up their own inference or wiring a different SDK per provider. Volunteer-run projects in particular cannot absorb surprise bills or off-topic misuse of a public chatbot. ## Solution [NYO](https://llm.nyolab.ai/) is "the governed AI gateway for agents and APIs", built by [Niyolab](https://nyolab.ai): one OpenAI-compatible endpoint and one API key to frontier open models (Kimi, GLM, DeepSeek and any OpenAI-compatible endpoint you register), per-key guardrails, prepaid credits with transparent per-model pricing (no subscriptions), request logs you can account for, a one-click "Add to Claude" integration, an MCP interface for agents, and bring-your-own-compute. ## Built on NYO ### Kuya J by BetterGov.ph "A civic assistant for government services, laws and budgets, answering in English and Filipino on open models through one key and one endpoint." BetterGov.ph is a volunteer-built, open-source portal (not an official government service). Kuya J grounds answers in BetterGov's own pages and Juris.ph legal data; guardrails turn away requests outside its civic purpose before they reach the model; NYO routes to stronger open models as they appear and spreads load across providers during spikes. - Try it: https://chat.bettergov.ph - BetterGov.ph: https://bettergov.ph - Full case study: https://llm.nyolab.ai/case-studies/bettergov ### Sappy by SapinSapin AI "A community project for Philippine languages answers questions about itself in Tagalog, Cebuano, Ilocano and Hiligaynon, through NYO." A community-manager agent deployed as a Cloudflare Worker that welcomes visitors and answers from SapinSapin's published datasets and model cards using a multilingual open-weight model via NYO. - Our case study: https://philippineaireport.com/case-studies/sapinsapin - Full case study: https://llm.nyolab.ai/case-studies/sapinsapin ### BetterCalatagan "An independent civic-information portal for Calatagan, Batangas, with a grounded assistant that answers in English, Filipino and Taglish from cited public records." Part of the BetterGov.ph / Better LGU community. Retrieval finds sourced passages, NYO (first route in a model chain with fallbacks) generates the answer, and application-level checks verify figures before display; a regression suite covers facts, finances, history, withheld items and prompt injection. 44 published pages, 25 barangay profiles, 8 open datasets, 838 finance checks across fourteen quarters. - Portal: https://bettercalatagan.vercel.app - Source: https://github.com/marcxxv/bettercalatagan - Full case study: https://llm.nyolab.ai/case-studies/bettercalatagan ### Valley O'Ville "An AI-native business operator: the front desk, the books and the follow-ups of a private family resort in Calatagan, carried by one person and an agent whose model calls go through NYO." A resort hosting up to thirty guests per stay; VOV Agent runs on glm-5.3-flash through NYO, handles guest inquiries on Facebook Messenger and the website chatbot, quotes, bookings, the ledger and social posts, with operator approval before anything is executed. One operator now covers what used to be three roles (front desk, bookkeeper, social media manager); 4.95-star guest rating; model cost of $0.15 per million input tokens and $0.50 per million output tokens — "the price of a few coffees a month". - Resort: https://valleyoville.com - Full case study: https://llm.nyolab.ai/case-studies/valleyoville Source and ongoing updates: https://llm.nyolab.ai/case-studies. Three of the four agents were built by [Marc Ocampo](https://marcocampo.com). ## Impact - One API key and one OpenAI-compatible endpoint replaces per-provider integrations - Per-key guardrails keep public civic bots on purpose and block off-topic requests before they reach a model - Four production agents across civic tech, language communities and hospitality, in English, Filipino, Taglish, Cebuano, Ilocano and Hiligaynon - A full small-business agent runs on prepaid credits at open-model prices (glm-5.3-flash: $0.15/M input, $0.50/M output tokens) - Open-weight models and bring-your-own-compute keep the stack sovereign and portable ## Philippine Relevance The 2025 Philippine AI Report found 65% of adopting organizations stuck at pilot stage and 57% blocked by talent scarcity. A governed gateway removes the infrastructure and procurement hurdle for exactly the builders who have the least slack: volunteer civic projects, language communities and MSMEs. Because the models are open-weight and the compute can be your own, it is also a practical path to sovereign Filipino AI. See also [Filipino LLMs](https://philippineaireport.com/llms) and the [AI startups directory](https://philippineaireport.com/startups). ## Disclosure NYO is built by Niyolab, whose founder Tim Santos is Editor in Chief of the Philippine AI Report. ## Links - Read the NYO case studies: https://llm.nyolab.ai/case-studies - Developer docs: https://llm.nyolab.ai/docs - Add your MSME agent: https://llm.nyolab.ai/msme