Integrations
Connect Claude Code, Open WebUI, n8n, LangChain, the DashScope and Ark SDKs, new-api and LobeChat to ModelSite: settings, a copy-paste check for each, and how each one was verified
Most tools only need a base URL and a key. This section covers the tools we ran for real against ModelSite (or read the source of), what to set, a minimal check you can paste, and the limits we found. For the full list of SDKs, frameworks and clients checked against official docs, see SDKs, Frameworks and Clients. For ComfyUI, see Use ModelSite in ComfyUI.
Status legend
| Status | Meaning |
|---|---|
| Run against ModelSite | We installed the tool, pointed it at ModelSite and made a real request. The page names the version and what came back. |
| Read from source | We read the tool's source code to confirm the paths, headers and fields it sends. We did not run it against ModelSite. |
What works
| Tool | What you can do | Status | Page |
|---|---|---|---|
| Claude Code | Coding with ModelSite chat models | Run against ModelSite (2.1.309) | Claude Code |
| Open WebUI | Chat, model list, image generation with Seedream | Run against ModelSite (0.12.3) | Open WebUI |
| n8n | Chat in workflows; video through the HTTP Request node | Run against ModelSite (2.44.6) | n8n |
| LangChain | ChatOpenAI and ChatAnthropic | Run against ModelSite (langchain-openai 1.7.2, langchain-anthropic 1.7.6) | LangChain |
| DashScope SDK, Ark SDK | Create Wan and Seedance video tasks and poll them | Run against ModelSite | Video SDKs |
| new-api | Use ModelSite as an upstream channel for Wan, HappyHorse and Seedance | Read from source (commit 92a09594) | new-api |
| LobeChat | Chat and Wan video through the Qwen provider | Read from source (2.2.7) | LobeChat |
Every example reads the key from the MODELSITE_API_KEY environment variable. A new key is sk-ms- followed by 64 hex characters. Create one in the console and never paste it into a shared file.
Not supported
/v1/videos(the OpenAI SDK'sclient.videos, and tools built on it) is not served. Video is served on each model family's own entrance; see the model pages under Video generation.callback_urlon video tasks is rejected with400. Poll the task instead.
Use ModelSite in ComfyUI
Two ways to generate video and images with ModelSite models in ComfyUI: ComfyUI's own API nodes pointed at ModelSite, or the ModelSite node pack, with what works, what does not, and how each was checked
Claude Code
Use ModelSite chat models in Claude Code: the three settings, the one thing to change for GLM models, and a one-line check