OpenClaw LaunchOpenClaw Launch
tool

SGLang

Serves your own language and multimodal models over an OpenAI-compatible API with high-throughput batching and prefix caching.

Install in ToolsView on GitHub ↗

What it provides

OpenAI-compatibleHigh throughputSelf-hosted models

Connect an HTTPS inference API. Model weights stay on CPU/GPU infrastructure and are never installed inside the 4 GB bot.

Connection requirements

An SGLang server started with `python -m sglang.launch_server --api-key ...` and published through HTTPS. Keep Bearer selected and paste the same key below. Weights, KV cache, and GPU memory stay on the inference host; the bot only installs an HTTP client.

Protocol
OpenAI-compatible REST
Endpoint
https://sglang.example.com
Authentication
bearer
Open official setup guide ↗

Where to add it

Open Dashboard Tools, select a running OpenClaw or Hermes instance, then use the card’s Install or Connect action. External services may require an HTTPS endpoint and credentials; the dashboard shows those fields before anything is saved.

Related Models & routing

OllamaDownloads, manages, and serves a broad catalog of local language and vision models through a simple API.llama.cppRuns quantized language models efficiently across CPU, Metal, CUDA, Vulkan, and other hardware backends.vLLMServes language models with high throughput, efficient memory use, batching, and OpenAI-compatible APIs.