星图

Astraflow 星图 is an Enterprise-oriented platform from UCloud (优刻得) for accessing and operating large language models without running your own training or inference stack. It bundles a unified model gateway with a sandboxed runtime for Autonomous Agents, positioning itself as a controllable foundation layer for company AI workloads rather than a single Standalone chatbot.

What it does

The platfORM brings together a large catalog of domestic and international models — including DeepSeek, Kimi, Qwen, GLM and MiniMax families — behind a single standardized API. Teams can switch between models quickly, send traffic across them, and connect them to APPlications through one entry point. AlongsIDE the model gateway, it offers a serverless sandbox for AI Agents that execute code or operate GUI desktops.

Key capabilities

• multi-model gateway: A single API and console for text, vision and multimodal models, allowing plug-in use and fast switching without separate deployments.
• Agent sandbox: A hardware-isolated runtime built on micro-VM architecture, with snapshot support for fast startup and state peRSIstence, aimed at compliance-sensitive deployments.
• Data infrastructure: Companion tools for analytics, search retrieval, backend development and long-term AI memory, useful for RAG and knowledge-base projects.
• Cost controls: Accounting per API key and per model, project- or team-level cost splitting, daily or monthly budget caps, quota alerts, throttling and abuse protection.
• SECurity and governance: API key organization management, role-based sub-accounts, model-level access blocking, IP allowlists and isolated test environments.

Who it is for

It suits enterprises building internal knowledge assistants, customer service bots or automated business agents, as well as engineering teams using strong Code Models for generation and debugging. Content producers exploring short-DRAMa or companion-style entertainment, and cross-border businesses needing regional compute and privacy compliance, are also target users. The platform claims a neutral stance, stating it does not touch customer data and does not compete with customer Businesses.

How to get started

Access is through the web console or the unified API. developers reportedly do not need to manage underlying compute scheduling or model deployment. New users may register, complete identity verification as required, create an API key and receive a free trial allowance — details such as the exACT token amount may vary and should be confirmed on the official site.

Pricing and access caveat

Pricing is usage-based and organized around tokens and model calls, with budget and quota management built in; a free trial exists alongside paid consumption, but rates and trial liMITs can change, so verify current terms before committing. The deeper compute layer, including GPU scheduling, is handled by a separate Ucloud platform. Availability, regional access and identity-verification requirements may differ by market, so check the provider's documentation for the latest conditions.