Open WebUI with Claude, GPT and Gemini

Add avots under Connections as an OpenAI API provider, and every avots model shows up in your Open WebUI model picker, billed on one balance.

Get your API key ✦ See pricing Docs on GitHub

Connect in three fields

Open WebUI talks to any OpenAI-compatible server. Add avots as a connection and its models appear in the picker.

FieldValue
Base URLhttps://api.avots.ai/openai/v1
API keyav_mcp_… · get your key
Modelanthropic/claude-sonnet-4.6 or any id from the table below

Setup

1. Open Connections

Go to Admin Panel → Settings → Connections, then under Manage OpenAI API Connections click the + button.

2. Paste URL and key

Set URL to https://api.avots.ai/openai/v1 and API Key to your av_mcp_ key.

3. Add a model id

In Model IDs (Filter) type anthropic/claude-sonnet-4.6 and click +. Add more ids the same way.

4. Save

Save the connection. The models show up in the Open WebUI model picker, billed on your avots balance.

Open WebUI · Admin Panel → Settings → Connections
URL:                 https://api.avots.ai/openai/v1
API Key:             av_mcp_YOUR_KEY
Model IDs (Filter):  anthropic/claude-sonnet-4.6

Open WebUI appends /chat/completions to the URL, so give it the /v1 root, not the full path. If the green connection check fails but chat works, ignore it and just add the model id by hand.

Example request

curl
curl https://api.avots.ai/openai/v1/chat/completions \
  -H "Authorization: Bearer av_mcp_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-sonnet-4.6",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

The same key also unlocks MCP tools

Your av_mcp_ key is not only for chat. avots also runs as an MCP server, so Open WebUI can call avots creative tools, image, video, music, talking avatars, on the same key and the same balance.

FieldValue
MCP serverhttps://mcp.avots.ai/
API keyav_mcp_… · same key as above

MCP setup guide →

Models and prices

Set the model field to any id below, or a lazy name like claude / gpt / gemini that maps to a sensible default.

ModelBest for⚡ / 1M tokens (in / out)
openai/gpt-5.5-proMax intelligence, hardest tasks60 000 / 360 000
anthropic/claude-opus-5Top Claude, long reasoning10 000 / 50 000
anthropic/claude-sonnet-4.6Balanced (default)6 000 / 30 000
google/gemini-3.1-pro-previewFresh Gemini Pro4 000 / 24 000
google/gemini-2.5-proStable Gemini Pro2 500 / 20 000
anthropic/claude-haiku-4.5Fast Claude2 000 / 10 000
deepseek/deepseek-r1Cheap reasoning1 400 / 5 000
google/gemini-2.5-flashFast and cheap600 / 5 000
openai/gpt-5-miniCheap OpenAI500 / 4 000
openai/gpt-oss-120b:freeFree (OpenAI open model, needs a positive balance)Free

* Price in ⚡ per 1,000,000 tokens, input / output (1000⚡ ≈ $1). Free models need a positive balance. Full list via GET /openai/v1/models.

Open WebUI with Claude, GPT and Gemini

Generate an av_mcp_ key in Settings, paste three fields, and reach 40+ models on one balance. The same key works for MCP.

Get your API key ✦
FAQ

Questions

Where do the models come from?
From the connection you add. Type the ids you want in Model IDs (Filter), or list them with GET /openai/v1/models.
Admin or per-user connection?
Either. The admin Connections panel shares it with all users; Direct Connections, if enabled, let each user add their own.
Why does the connection check fail?
Open WebUI verifies some providers strictly. Chat still works, just add the model id manually.
How is it billed?
Per request from your avots balance, the same wallet as the web app and Telegram.
Does it support function calling and tools?
Yes. The endpoint forwards the standard OpenAI tools, tool_choice and parallel_tool_calls fields (and the legacy functions/function_call), in both streaming and non-streaming mode, so agent frameworks like LangChain, LlamaIndex, Langflow and n8n get real tool_calls back. Whether a given request uses tools depends on the model you pick: Claude, GPT and Gemini all support function calling.

Other clients

OpenAI-compatible API

avots speaks the OpenAI API. Any app that talks to ChatGPT through base URL, key and model…

LibreChat with Claude, GPT and Gemini

Add one custom endpoint block to librechat.yaml and avots appears as a provider with every…

Cursor with Claude, GPT and Gemini

Override the OpenAI base URL in Cursor Settings, Models, paste your av_mcp_ key, add a mod…

Python openai SDK with avots

Pass base_url and your av_mcp_ key to the official openai Python client, then call any mod…

openclaw with Claude, GPT and Gemini

Add an OpenAI-compatible provider to openclaw.json pointing at avots, and openclaw runs an…

Hermes with Claude, GPT and Gemini

Set the OpenAI base URL and your av_mcp_ key in Hermes, pick a model, and it runs on avots…

Cline with Claude, GPT and Gemini

Pick OpenAI Compatible in Cline settings, paste the avots base URL and your av_mcp_ key, t…

Continue.dev with Claude, GPT and Gemini

Add one models entry to Continue config.yaml with provider openai and the avots base URL, …

Aider with Claude, GPT and Gemini

Export two environment variables and launch Aider with an openai/ model prefix, and the CL…

SillyTavern with Claude, GPT and Gemini

Choose Custom (OpenAI-compatible) as the Chat Completion source in SillyTavern, paste the …

AnythingLLM with Claude, GPT and Gemini

Choose Generic OpenAI as the LLM provider in AnythingLLM, fill the avots base URL, key, mo…

Node.js openai SDK with avots

Pass baseURL and your av_mcp_ key to the official openai client, then call any model by id…

Chatbox with Claude, GPT and Gemini

Add a custom OpenAI API compatible provider in Chatbox, paste the avots host and your av_m…

Get your API key