AICraft

AICraft is a multi-model API aggregation platform designed for developers and Enterprises who need to access dozens of large language models through a single, OpenAI-compatible endpoint. Instead of manAGIng multiple API keys, SDKs, and billing systems from different model provIDErs, AICraft consolidates everything into one unified gateway. Its core innovation is the Auto Router, which automatically identifies the type of task (e.g., coding, translation, reasoning, creative writing) and routes the request to the most cost-effective model, with automatic failover if the primary model is unavailable. A dual-layer caching system (semantic cache and provider cache) further reduces latency and token costs for repeated or similar queries.

What it does

AICraft ACTs as a middleware layer between your APPlication and 29 upstream model providers, offering 88+ models including both domestic Chinese and international LLMs. You keep a single API key and a single codebase; just change the base URL in your existing openai-compatible code to point to AICraft. The platfORM handles authentication, format adaptation, Intelligent routing, fault tolerance, and billing aggregation. It also provides a web console for browsing the model catalog, Testing requests, and viewing usage logs and invoices.

Key capabilities

• Unified API access: Fully compatible with the OpenAI chat Completions protocol. Use any SDK (Python, Node, Go, Rust, cURL) with minimal code changes.
• Auto Router: Automatically detects task type (programming, translation, Chinese generation, reasoning, Creative writing, customer service) and selects the best model. Remembers usage preferences and switches to a backup model on failure.
• Dual-layer caching: Semantic cache returns cached results for semantically similar requests; provider cache reuses upstream cache. Reduces Token Consumption and response time for repetitive queries.
• Team management: enterprise plans support sub-keys with quota isolation, team member management, audit logs, invoices, and SLA contracts.
• Data Compliance: TLS 1.3 encryption, hashed API keys, no stoRAGe of full conversation content. Domestic models keep data within China; overseas models only transMIT anonymized content, complying with PIPL, GDPR, CCPA, and LGPD.

Who it is for

AICraft is best suited for AI application developers, AI agent teams, and small-to-medium businesses that need to integrate multiple LLMs without the overhead of managing separate integrations. It is also ideal for technical teams performing model comparison and selection, or those requiring sub-key management and unified billing. IndiVidual developers can use the free trial credits to experiment with various models quickly.

How to get started

Register an account on the AICraft website to receive a free trial credit (5 CNY). Obtain your API key, then modify your existing OpenAI-compatible code by changing the base URL to AICraft's endpoint. Set the model parameter to "auto" to enable smart routing, or specify a model ID from the catalog. Monitor usage and costs through the web console. For enterprise features, upgrade to a paid plan that matches your concurrency needs.

Pricing and access caveat

AICraft operates on a prepaid, pay-as-you-go model. New users get a small Free trial credit. Paid plans start at any recharge amount, with higher tiers offering increased concurrency limits (from 20 to 300 concurrent requests). Enterprise plans include sub-keys, team management, invoices, and SLA guarantees. Note that the gateway introduces minimal additional latency (documented as ~0.3ms routing overhead); for ultra-low-latency scenariOS, direct provider access may be preferable. The Auto Router works well for common tasks but may not suit highly specialized use cases where manual model selection is recommended.