Openclaw Model Bridge

openclaw-model-bridge

**Agent Runtime Control Plane** — Connect any LLM to [OpenClaw](https://github.com/openclaw/openclaw) with one command. Zero dependencies, **12 providers** (含豆包 + DeepSeek-V4-Pro 量化/满血双端点 + GLM-5.2 coding), multimodal support, reasoning capability. 将任意大模型（Qwen / OpenAI / Gemini / Claude / Kimi / MiniMax / GLM / Doubao Seed 2.0 / **DeepSeek-V4-Pro**）一键接入 OpenClaw — 零第三方依赖、支持多模态、10 分钟跑通。

[](https://opensource.org/licenses/MIT) [](https://www.python.org/downloads/) []() []() []() []() []() []() []()

**Current version:** `v37.9.330` / `0.37.9.159` (2026-08-25) — see [`CLAUDE.md`](CLAUDE.md) for full changelog. **Latest milestone:** 📄 **ArXiv paper published** — [**arXiv:2606.14589**](https://arxiv.org/abs/2606.14589) (2026-06-15, cs.SE): the *fail-plausible* concept + a 5-class taxonomy of silent failures from 22 production incident postmortems (IEEE Software in review; ISSRE rejected 2026-08-14, conference track closed — journal-first strategy V37.9.307). **Constitutional priority now: LLM-Observer (机械化人眼)** — an automated user-perspective observer that catches *fail-plausible* silent failures **before the user does** (the paper's headline open problem: ~70% of silent failures were caught by looking at the product, while tests/