Project Phoenix is yGen's end-to-end AI agent platform — deployed on the AI Agent Box, running entirely on your premises, under your jurisdiction. Multi-agent teams. Sandboxed code execution. Built-in RAG. Zero data egress. The AI workforce your organisation needs, without sending a single token to the cloud.
Every major cloud vendor — Anthropic, OpenAI, AWS — now offers an AI agent platform. They are excellent. They are also, by design, not in your building. Every prompt crosses a border. Every token is subject to foreign jurisdiction. For hospitals, banks, law firms, insurers, and government agencies in the Philippines and across APAC, that is not a feature trade-off — it is a regulatory disqualifier. Phoenix runs where the cloud cannot go: inside your perimeter, on hardware you own, governed by rules you control.
Phoenix is not a chatbot builder or a prompt runner. It is a full-stack AI agent operating system — built to deploy autonomous, multi-agent teams that think, delegate, execute, remember, and ship.
The AI decides at runtime which specialist to call, which tool to use, and which skill to run — no pre-wired decision trees, no scripted paths.
Build an orchestrator that delegates to specialist sub-agents — each with its own model, prompt, knowledge base, and toolkit. Mirror the structure of a real team.
Every conversation runs inside its own isolated Docker container with a real terminal, file tree, and live logs. Agents write code and run it — safely.
Drag-and-drop agent design on a React Flow canvas. Non-technical teams build visually; the AI runtime executes with production-grade reliability underneath.
Native adapters for Telegram, Slack, Discord, and WhatsApp. Point your agent at a channel and it is live. Per-channel conversation memory included.
Upload PDFs, DOCX, CSV, Markdown. Auto-chunked, embedded via Ollama, stored in Qdrant. Agents search your private knowledge base with semantic search and cited answers.
Conversations are checkpointed to PostgreSQL. Agents remember context, files, and decisions across sessions — turning transactional bots into long-term assistants.
Agents gain capabilities through portable Skill packages and Model Context Protocol servers — the open standard for agent interoperability. First published: @ygen-phoenix/rag-mcp.
Live dashboard showing agent activity, tool calls, error rates, and streaming event logs per run. Drill into any tool call payload with the built-in JSON explorer.
Use Llama, Mistral, Qwen on-box via Ollama. Optionally route to Claude, GPT-4o, or Gemini when permitted. Switch models without rewriting flows.
An in-app deepagent creates, configures, and enhances agents from plain conversation. "Build me a claims processing agent" → working agent in one turn. No forms. No code. No delay.
Generate a script snippet and paste it into any website. A floating chat widget delivers your agent to any page — no Phoenix login required for end users. Live in five minutes.
Anthropic, OpenAI, and AWS have excellent agent platforms. They are also, structurally, subject to foreign jurisdiction, unpredictable unit economics, and the requirement that your most sensitive data leaves your building. Phoenix was engineered around those three constraints.
Every inference call, every RAG query, every agent memory read — all happen inside your perimeter. No prompt, no document, no customer record ever touches the internet. Air-gap deployment is supported for SCIF and classified environments.
Per-token cloud pricing is a feature when usage is light and a liability when it isn't. A 24/7 claims agent or legal review pipeline burns predictable token volume. Phoenix turns that into a flat operating cost — hardware capex amortized over years, not a monthly bill that grows with success.
BSP, MAS, OJK, PDPC, BNM, DICT. Phoenix is built in the Philippines, governed from Singapore, and designed from day one around APAC regulatory requirements — not retrofitted from a US-first compliance posture. Your auditor can walk to the server and inspect it.
Open standards throughout. Ollama for inference — swap models without rewriting flows. Qdrant for vector storage. PostgreSQL for memory. MCP for tool interoperability. If you choose to migrate, you own your data, your models, and your agent configurations.
Every regulated enterprise in the Philippines and across APAC that processes sensitive data — financial records, patient files, legal documents, government data — faces the same structural barrier to cloud AI adoption. Phoenix was built to be the answer.
Patient triage, record analysis, clinical decision support — on-premise by regulation.
Contract review, due diligence, precedent RAG — attorney-client privilege demands air-gap.
Credit scoring, fraud detection, relationship banking — BSP and MAS compliance required.
Claims processing, actuarial modeling, fraud analysis — sensitive data stays on-prem.
Citizen services, internal automation, document processing — DICT-governed data sovereignty.
Academic advisors, curriculum RAG, student support — private institution data control.
Operations monitoring, predictive maintenance, supply chain — industrial data stays local.
Lead intelligence, property RAG, client relationship agents — PH's largest private sector.
There is no other on-premise AI agent platform purpose-built for APAC regulated markets. The cloud players cannot move on-prem. Legacy on-prem vendors cannot move to agents. Phoenix occupies the intersection alone.
Cloud platforms are racing to add generic capabilities. Phoenix is going the opposite direction — deeper into flagship verticals with purpose-built agent packs, RAG templates, and pre-tuned workflows. That depth is a durable moat.
One-time hardware sale paired with mandatory annual support and license creates a high-margin recurring revenue stream from Year 2. Every box shipped is a revenue stream that compounds — not a one-time transaction.
Whether you are evaluating a partnership, a deployment, a consulting engagement, or an investment — here is what Phoenix means for you specifically.
If you are a systems integrator, IT reseller, or solutions provider serving banks, hospitals, government, or law firms — your clients are asking about AI and hitting a wall: their data cannot leave the building. Phoenix is the answer you have been waiting to offer.
Phoenix runs in your server room, under your IT policy, auditable by your team. Your data never moves. Your regulators never object. And your operations team gets a multi-agent AI workforce that works 24/7 without headcount.
The AI agent market is growing fast and consolidating around cloud platforms. That consolidation is creating a structural void — the regulated, sovereignty-sensitive enterprise segment that cannot adopt cloud AI. Phoenix is the only platform purpose-built to fill that void in APAC.
Most AI transformation engagements stall at the data governance layer — not because the AI isn't good enough, but because it can't clear legal and compliance. Phoenix removes that blocker, giving you a deployable platform to anchor your AI roadmap recommendations.
If your data has to stay in-country, your model has to run under your roof, and your operations have to scale without growing headcount — that is exactly what Phoenix was built for. Let's talk about what a pilot deployment looks like for your organisation.