XCUBEby Enlight CorpX Cube is an enterprise private AI platform: your data stays in your environment, and you control the compute.
Employees open a browser to chat, search documents, and organize work — no code needed.
Products already on the OpenAI SDK just point the base URL at X Cube.
Upload SOPs, specs, and contracts; answers come from company knowledge and cite it.
42 on-prem and 5 cloud models, chosen by cost, speed, and data governance.
Sensitive data and inference stay on your own NVIDIA DGX Spark / GB10 nodes.
Every key has a budget, rate limit, and model allowlist, and can be frozen; sending or paying always asks you first.
POST /v1/chat/completions The same models and settings your people use.
/llms.txt/v1/openapi.json/mcp Read once and know how to connect.
/v1/rag/query Citations come back, so agents work from what people see.
/v1/catalog Check models and their lifecycle without a key, and skip the ones retiring soon.
On-prem and cloud speak the same API format, so changing deployment changes nothing for callers.
402429403 402: stop and tell a human. 429: back off. 403: switch models.
People and agents share the same models, the same knowledge, and the same permissions — so what agents use and spend shows up in the same console people already manage.
Enlight Corp helps enterprises bring an AI platform and AI agents into their own environment — chat, APIs, open models, and on-prem GB10, with data kept inside.
Teams with an account sign in directly; for new companies we first confirm the deployment model and data scope.
Paste this line into ChatGPT, Claude, or your AI assistant. It compares X Cube with your needs, lists where it fits, and drafts an inquiry you decide whether to send.
X Cube gives employees a direct workspace and gives engineers the same AI capability through APIs.
Teams can choose models based on cost, performance, data governance, and language quality.
When data needs to remain in your environment, X Cube can move toward DGX Spark-based local deployment.
X Cube connects chat, APIs, company knowledge, open models, and DGX Spark local nodes into one rollout path.
Once deployed, products already using the OpenAI SDK only change the base URL and key. On-prem, the base URL is your own host.
# List the models your organization can use curl https://<your-xcube-host>/v1/models \ -H "Authorization: Bearer $XCUBE_API_KEY" # Python — change base_url only client = OpenAI( base_url="https://<your-xcube-host>/v1", api_key=os.environ["XCUBE_API_KEY"], )
Frontier models, agent protocols, and enterprise AI platform practice. Every article carries its source, a key answer you can quote, and a Markdown copy for AI assistants.
Agent onboarding is the path that lets an AI agent read, evaluate, and connect to a product on its own: llms.txt and Markdown versions make content readable, a restricted mode or human approval controls credentials, and people can see what the agent will do and what it must ask them first.
Enterprise agent autonomy should be designed around the human's role: low-risk actions such as reading, searching, and drafting run automatically; irreversible or external actions such as sending, paying, or writing to outside systems need human approval; and forbidden actions are listed explicitly. Research shows trust is calibrated per task and that GUI agents are easily misled by dark patterns.
X Cube is an enterprise private AI platform with team chat, an OpenAI-compatible API, knowledge-base Q&A, and open models. It runs in the cloud or on NVIDIA DGX Spark / GB10 nodes on-prem, keeping data in an environment the company controls.
With an on-prem deployment, inference runs on your own GB10 nodes and data stays inside. If you use cloud models, those requests go to the model provider; admins can restrict which models the organization may use.
Yes. Point the base URL at /v1 on your X Cube host and authenticate with an sk-xcube API key. Chat completions, completions, embeddings, responses (subset), and knowledge-base queries are supported.
As of 2026-10-04, the public catalog lists 5 cloud models (GLM-5.1, GPT-4o, GPT-4o mini, Claude Sonnet 4.6, Claude Opus 4.7) and 42 available on-prem models, including Qwen3, DeepSeek V4, GLM-5.2, gpt-oss, Whisper, and FLUX. See /v1/catalog for the full list.
Yes. Once X Cube is deployed in your environment, agents reach internal tools through MCP, agent runs can require human approval or rejection, and every API key has a budget, rate limit, model allowlist, and freeze. To check fit first, ask an AI assistant to read for-agents.md; it lists where X Cube fits and drafts an inquiry. See how an AI assistant evaluates →
Teams with an account sign in to X Cube Chat. Companies new to X Cube email ai@enlightcorp.com.tw to book a consultation; we first confirm the deployment model (cloud or on-prem) and data scope.