On September 2, 2026, Moonshot AI announced that its Kimi API now supports both the OpenAI Responses API format (base_url: https://api.moonshot.cn/v1) and the Anthropic Messages API format (base_url: https://api.moonshot.cn/anthropic). This update is specifically designed for users of OpenAI’s Codex and Anthropic’s Claude Code, allowing them to connect directly to Kimi’s models—including kimi-k3 and kimi-k2.7-code-highspeed—without the need for format conversion or a local proxy setup.
The newly added compatibility means developers and enterprises using these popular coding assistants can now seamlessly switch to Kimi’s backend infrastructure while retaining their existing workflows and client configurations. By eliminating the overhead of translating API requests between different providers, Moonshot AI aims to lower the barrier for teams testing or migrating to its high-performance models.
In terms of input capabilities, the Kimi API supports both text and image inputs across these compatible endpoints. However, video input is not currently supported; users who need to analyze video content must first extract key frames before sending them through the API. This limitation is clearly documented to help developers plan their integration accordingly.
To assist with the transition, Moonshot AI has published comprehensive configuration guides and reference documentation on its official developer portal. These materials detail the exact base URLs, authentication methods, request/response schemas, and practical examples for both the OpenAI-compatible and Anthropic-compatible paths. The documentation also covers known differences in behavior, error handling, and rate limits compared to the original providers.
This strategic move reflects a broader industry trend toward interoperability among AI model providers. As large language models become more specialized and cost-effective, offering drop-in compatibility with dominant API ecosystems is seen as a key differentiator. For Moonshot AI, which has been expanding its footprint in both Chinese and international markets, this update is likely to accelerate adoption among developers who have standardized on OpenAI or Anthropic tooling.
The kimi-k3 and kimi-k2.7-code-highspeed models, which are now accessible through these new endpoints, are positioned as competitive options for code generation, reasoning tasks, and rapid prototyping. By supporting both major API standards simultaneously, Moonshot AI gives its users the flexibility to integrate Kimi into existing pipelines with minimal disruption, while also opening the door for future cross-provider collaborations.
