EN ▾
Get API key

Open WebUI Anthropic API vs an Uncensored Model

Updated

Open WebUI can hold several model connections at once, so you do not have to choose one provider forever. This page covers adding an Anthropic model, adding ours next to it and picking per chat.

A control panel with two connections feeding one chat window, illustrating two model backends in one interface

Key points

  1. Open WebUI talks to OpenAI-compatible endpoints; our API needs only a URL, a key and the model ID uncensored.
  2. For Anthropic models, follow the current Open WebUI and Anthropic docs, since the connection method is theirs and can change.
  3. Use the Anthropic model for work that suits its policies and an uncensored one for lawful adult fiction it will not write.
  4. Our side costs $0.15 per 1M input and $0.50 per 1M output tokens; a 3,000 token prompt with a 500 token reply is $0.0007.

Two connections, one interface

If you searched for open webui anthropic api, you probably want Claude models inside the same chat window you already use for everything else. That works, and it does not exclude other backends. Open WebUI lets you add more than one connection and then pick the model from a dropdown at the start of each chat. So the real question is not Anthropic or something else, it is which model gets which job.

This page does three things. It shows how to connect an OpenAI-compatible endpoint, using ours as the worked example. It describes how Anthropic models fit into Open WebUI at a general level without pretending to replace their documentation. And it gives a plain decision rule for when to reach for each, with costs for our side calculated exactly.

Connecting Anthropic models

We do not operate the Anthropic API and will not tell you its prices or limits. Check their documentation and your account for current figures. What can be said safely is how the pieces relate.

Open WebUI's native connection type speaks the OpenAI chat completions format. Anthropic publishes its own native API, and at the time of writing there are two common ways to bring its models in: a community function or pipe installed inside Open WebUI, or an OpenAI-format compatibility route that Anthropic documents for its API. Both change over time, so open the current Open WebUI documentation for connections and the current Anthropic documentation for compatibility, and follow whichever is recommended today.

Whichever route you take, keep the Anthropic key separate from every other key, and set a spending limit in your Anthropic account if one is offered. Hosted frontier models apply their own content policies to every request, so read those before you plan creative work around them.

Adding an OpenAI-compatible endpoint

This part is simple because it is the connection type Open WebUI was built around. In the admin settings, open Connections, add a new OpenAI API connection and enter:

  • URL: https://api.nsfwwan.cc/v1
  • Key: your API key

Save, and the model list refreshes by calling GET /v1/models. You should see uncensored appear. Labels differ slightly between Open WebUI releases, but the fields are always a URL and a key.

If you start Open WebUI with Docker, you can supply the connection through environment variables instead:

docker run -d -p 3000:8080 \
  -e OPENAI_API_BASE_URL="https://api.nsfwwan.cc/v1" \
  -e OPENAI_API_KEY="YOUR_KEY" \
  -v open-webui:/app/backend/data \
  --name open-webui \
  ghcr.io/open-webui/open-webui:main

Before you do either, confirm the key works from a terminal. This also shows the headers you will rely on later:

curl -i https://api.nsfwwan.cc/v1/chat/completions \
  -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"uncensored","messages":[{"role":"user","content":"Say hello in five words."}],"max_tokens":30}'

Look for X-Balance-USD in the response headers. It shows the credit that remains. To get a key, use the key page; new users can claim a $0.50 trial there with one click, valid for 7 days and without payment details.

Which model for which job

TaskBetter fit
Code review, long document analysis, careful reasoningAn Anthropic model, if its quality is what you need
Anything where you need a named vendor model for compliance or consistencyThat vendor, directly
Fiction, roleplay and companion chats for lawful adult use that mainstream models declineThe uncensored endpoint
High-volume drafts where per-token price mattersCompare current prices, then test both on your prompts
Images, audio, embeddingsNeither here; our API is text only

We are a small text-only service. Our model is one open-weight model run by an independent provider, tuned to avoid content refusals. It is not Anthropic's model and does not claim to match it on every task. If Claude suits your work and its policies, use it. The uncensored endpoint is for the cases where a refusal stops the work.

One rule stays fixed on our side: no sexual content involving minors, regardless of framing.

A practical pattern is to keep the Anthropic model as the default for daily work and switch to the uncensored one only when you start a creative session. That keeps costs predictable, since you only pay our per-token rate for the chats that need it. It also keeps your history tidy, because fiction chats and work chats live in separate threads with clearly different model names. If you later find that one backend covers all your needs, removing the other connection takes a minute and loses nothing.

What the uncensored side costs

Our price is $0.15 per 1M input tokens and $0.50 per 1M output tokens, billed from prepaid credit. Failed or refused requests are not charged, and credit never expires. Two examples in Open WebUI:

A short chat message

A 3,000 token prompt (system text plus history) and a 500 token reply:

  • Input: 3,000 / 1,000,000 x $0.15 = $0.00045
  • Output: 500 / 1,000,000 x $0.50 = $0.00025
  • Total: $0.0007

A thousand of these cost $0.70.

A long document in the context

Upload a 60,000 token manuscript and ask for a 1,000 token critique. Input is 60,000 / 1,000,000 x $0.15 = $0.009, output is 1,000 / 1,000,000 x $0.50 = $0.0005, for $0.0095 in total. The window is 100,000 tokens for prompt plus reply, so this fits, but follow-up questions resend the document each time and each costs about the same again. Open WebUI's retrieval feature can send only relevant parts, which cuts the figure.

Top-ups run from $20 to $500 in whole dollars with a +5% bonus from $50, +10% from $100 and +15% from $200. This site takes crypto only, using USDT on TRC20 or USDC on Base. See the pricing page.

If the connection does not work

Most failures come down to a handful of causes, and you can rule them out in order.

  1. The URL. It must end at /v1. Do not add /chat/completions yourself, because Open WebUI adds the path it needs. A missing /v1 or a doubled path gives a not found error.
  2. The key. Paste it without spaces or quotes. A 401 response means the key was rejected, and a fresh copy from your account page settles it.
  3. The model list. If the dropdown is empty, run the GET /v1/models check from a terminal. If curl works and Open WebUI does not, the problem is in the connection form, not the service.
  4. Verification. Some Open WebUI versions have a verify button next to the connection. Use it after each edit instead of starting a chat to test.
  5. Long answers that stop early. Check any max tokens setting in the chat controls. The API itself does not require one, and the completion can use the rest of the 100,000 token window.

When you contact support, quote the X-Request-Id response header. It identifies the exact request on our side and saves a round of questions.

Practical tips for running both

  • Name your models. In Open WebUI you can rename and describe models in the admin panel, so the dropdown makes clear which one is which.
  • Keep system prompts separate. A prompt tuned for one model often reads oddly to another. Create presets per model.
  • Do not switch models mid-chat for sensitive work. Each model sees the entire history. Decide the backend before you paste anything private.
  • Watch concurrency. A full key allows 8 concurrent requests and 300 per minute; the trial key allows 2 and 60. If several people share one Open WebUI, a 429 with code concurrency means too many streams are open at once. A 429 with a Retry-After: 5 header is temporary overload, so retry shortly.
  • Streaming works. With streaming on, the final chunk includes usage, which helps you track spend.

Prompts sent to our API are not used for training. More detail on parameters and tool calling is in the quickstart.

Questions and answers

Can Open WebUI use the Anthropic API directly?

Open WebUI documents the supported ways to add Anthropic models, such as functions or a compatibility route. Check its current docs and Anthropic's, because these methods can change.

What model ID do I use for your endpoint?

Use uncensored. The base URL is https://api.nsfwwan.cc/v1 and authentication is a Bearer key.

Can I keep both connections active at once?

Yes. Open WebUI shows models from every connection in one dropdown, so you pick per chat.

Do you offer Anthropic models?

No. We serve one open-weight model run by an independent provider. We are not the Anthropic API and not a router across vendors.

Is there a way to try it without paying?

New users can claim $0.50 of trial credit with one click on the key page. It lasts 7 days, needs no payment details and is one per person.

Your key is one form away

Create an account, copy the key, change the base URL. That is the whole setup.

Get API key