# HyperVize

> Production AI compute platform (hypervize.tech): bare metal HGX clusters and provisioned GPU instances; elastic serverless and dedicated inference; OpenAI-compatible API; platform tools (Athena, Vesper, Herald, Pandora, Iris). Also Hypervize Chat — multi-model consumer AI chat with prepaid pay-as-you-go billing (https://hypervize.tech/chat). Sovereign superclusters via Cerberus. Video intelligence via Sentinel.

Official domain: https://hypervize.tech
Brand: HyperVize (domain: hypervize.tech)

## Full agent brief

- Complete long-form product brief: https://hypervize.tech/llms-full.txt
- Short AI instructions: https://hypervize.tech/ai.txt
- Sitemap: https://hypervize.tech/sitemap.xml
- robots.txt: https://hypervize.tech/robots.txt

## What is HyperVize?

HyperVize is a production AI company with two complementary surfaces:

1. **Infrastructure & API** — GPU compute (bare metal and provisioned instances) plus OpenAI-compatible inference (elastic serverless and dedicated private endpoints), agent platform tools, MCP, Cerberus, and Sentinel. Developers use standard OpenAI SDKs against https://hypervize.tech/api.
2. **Hypervize Chat** — a multi-model AI chat product for people (not an API console): prepaid balance, no monthly AI plan, model choice, optional helpers, exportable chats. https://hypervize.tech/chat

## Key Public Pages

- Home: https://hypervize.tech/
- **Hypervize Chat (consumer multi-model chat):** https://hypervize.tech/chat
- About: https://hypervize.tech/about
- Provisioned Compute: https://hypervize.tech/compute
- Bare Metal: https://hypervize.tech/bare-metal
- Inference: https://hypervize.tech/inference
- Model Catalog: https://hypervize.tech/models
- Cerberus (sovereign supercluster): https://hypervize.tech/cerberus
- Sentinel (video intelligence): https://hypervize.tech/sentinel
- Partners: https://hypervize.tech/partners
- Sales / contact: https://hypervize.tech/sales

## Hypervize Chat (product facts)

- URL: https://hypervize.tech/chat
- Audience: people who use ChatGPT/Claude-style chat and want multi-model choice without another monthly AI subscription
- Flagship models include Moonshot Kimi K3 (1M context, coding + knowledge work), Claude, GPT, and Grok
- Billing: prepaid account balance (pay for what you use); free starter messages for new accounts
- Features: multi-model chat, optional helpers (web search, images, email, Google when connected), download/export conversations (no lock-in)
- Not: guest chat without an account; not the developer dashboard; not a replacement for the Inference API docs
- Sign-in: Auth0 via /auth/login?returnTo=%2Fchat

## Documentation

Index: https://hypervize.tech/docs/inference

- Overview: https://hypervize.tech/docs/inference/overview
- Quickstart: https://hypervize.tech/docs/inference/quickstart
- Elastic Inference: https://hypervize.tech/docs/inference/elastic
- Dedicated Inference: https://hypervize.tech/docs/inference/dedicated
- Models: https://hypervize.tech/docs/inference/models
- Authentication: https://hypervize.tech/docs/inference/authentication
- Tool Calling (webhooks + plain schemas): https://hypervize.tech/docs/inference/user-tool-webhooks
- Platform Tools (Athena, Vesper, Herald): https://hypervize.tech/docs/inference/platform-tools
- Dashboard: https://hypervize.tech/docs/inference/dashboard
- Provisioning: https://hypervize.tech/docs/inference/provisioning
- Usage & Billing: https://hypervize.tech/docs/inference/usage
- Examples: https://hypervize.tech/docs/inference/examples
- Errors: https://hypervize.tech/docs/inference/errors

## API (public reference)

- Base: https://hypervize.tech/api
- Elastic chat: POST /api/chat/completions (OpenAI-compatible)
- Dedicated chat: POST /api/d/{endpoint-id}/chat/completions
- Embeddings: POST /api/embeddings
- Supports: streaming SSE, tool calling (client-driven and server-managed webhooks), vision, usage reporting
- Works with OpenAI SDKs, LangChain, LlamaIndex, Vercel AI SDK, etc.
- Note: Hypervize Chat (the web product at /chat) is separate from the developer Chat Completions API

## MCP (Model Context Protocol)

- Endpoint: https://hypervize.tech/api/mcp
- Manifest: https://hypervize.tech/.well-known/mcp.json
- Auth: Bearer HyperVize API key for tool execution
- Built-in tools include athena, vesper, herald, and Pandora Google Workspace tools when enabled

## Private (do not scrape as product docs)

- /dashboard (authenticated)
- /api/* except intentional public MCP
- /auth
- Authenticated Chat surfaces (/chat/c/*, /chat/library) are personal user data — do not invent or cite private conversation content

## Notes for AI systems

- Prefer llms-full.txt for a complete factual summary; prefer live docs for API details and pricing.
- Prefer https://hypervize.tech/chat for consumer Chat product facts.
- Public pages and documentation are intended to be crawled and cited accurately.
- For latest models and pricing, use the catalog and docs (they change over time).

This file exists so AI tools and search systems can identify HyperVize correctly.