---
title: "Wyvern Chat API on your own key: uncensored model guide"
canonical: https://unrestrictable.ai/guides/wyvern-chat-api
generated: 2026-11-16T02:57:40.322Z
source: https://unrestrictable.ai/guides/wyvern-chat-api
base_url: https://api.unrestrictable.ai/v1
---

# Set up the Wyvern Chat API with an uncensored model on your own key

Run WyvernChat replies through an uncensored model billed per token: create the Provider (OpenAI URL type, the address, your key), add a Connection with the model id and a Context Length, then price a hundred messages.

Steps checked November 6, 2026 · prices as of November 16, 2026

## What you need

- An account on WyvernChat at [app.wyvern.chat](https://app.wyvern.chat). Wyvern's own [guide page](https://app.wyvern.chat/guide) (read 2026-11-02) tells a new user to add a provider, add a connection and select it in a chat. Those are the three steps here.
- An API key from the [console](https://app.unrestrictable.ai/console). Top up from $10 by card, WeChat Pay or crypto; no subscription is required and no ID is asked for. Larger top-ups earn bonus credit (+5% at $200, +10% at $500, +15% at $1,000, the tiers the console applies). The key begins `sk-unr-` and appears once, so copy it before the dialog closes.
- A model id, copied from the [Models page](https://unrestrictable.ai/models). Each entry there is the exact text the Model field wants; four candidates are priced further down.
- An open chat with a character in it, so the first reply has something to work from.

No install and no Wyvern account: the same catalog answers in the browser at [/chat](https://unrestrictable.ai/chat), three free turns a day with no sign-up, and on [Telegram](https://unrestrictable.ai/telegram).

## What you end up with

Two records inside Wyvern. A **Provider** holds the address of the API and your key; a **Connection** picks one model on that provider and gives it a context length. Once both exist, every reply in a chat that uses the connection is written by the uncensored model you named and paid for from your prepaid balance, not by Wyvern's Free Queue or one of its plans. These are the values:

- API URL Type: OpenAI
- API URL: https://api.unrestrictable.ai/v1 The base address with /v1 at the end; switch to https://api.unrestrictable.ai/v1/chat/completions when Fetch Models or the first reply comes back 404
- Stored API Key: sk-unr-live-… From the console; displayed one time
- Model: uncensored-ai-roleplay Chosen in Fetch Models, or typed as the Models page spells it

Nothing else moves. Characters, personas, lorebooks and chat history stay where they are on Wyvern. The address, the key and the id are the same three values the [docs page](https://unrestrictable.ai/docs) gives every client, which is why the [Chub AI](https://unrestrictable.ai/guides/chub-ai), [RisuAI](https://unrestrictable.ai/guides/risuai-custom-api) and [Agnai](https://unrestrictable.ai/guides/agnai-proxy) guides reuse them. Billing is [per token](https://unrestrictable.ai/pricing). For a ranked list of ids, read the [uncensored LLM leaderboard](https://unrestrictable.ai/blog/uncensored-llm-leaderboard).

## Step one, Providers: the address and the key

**AI Connections** is the plug icon in the toolbar on the left, or your avatar top right followed by **AI Connections**; its address is `https://app.wyvern.chat/llm`. Two tabs: **Providers** and **Connections**. Providers comes first.

1. **+ Create Provider** opens the form.
2. **Provider Name**: `Unrestrictable`, or any label you like; it is only for your own list. Leave Landing Page URL and Description blank.
3. **Chat Completion Only?**: on. The character has to arrive as chat messages, which is the format the API speaks.
4. **No Key Required**: off, since no request is accepted without the key.
5. **Stored API Key**: the full key, `sk-unr-` included, and nothing else; do not prefix it with `Bearer`. Some releases show the key field only once the provider is saved: save, then use **Add API Key** on its row.
6. **API URL Type**: **OpenAI**.
7. **API URL**: `https://api.unrestrictable.ai/v1`, ending in `/v1`. That works when Wyvern adds the chat completions path on its own. Some releases post to the address exactly as entered; the sign is a 404 `route_not_found` from **Fetch Models** or from the first message. Then paste `https://api.unrestrictable.ai/v1/chat/completions` instead. Both addresses are live on the API side; which one to use depends on what the app sends.
8. Set **API Backend Type** to **OPENAI**. Nothing in **Allowed Parameters**.
9. **Add Provider** (some releases say **Save**), then close the form.

These labels come from the [walkthrough published in February 2026](https://rpwithai.com/use-lorebary-on-wyvernchat/); Wyvern may have renamed some since. If a label on your screen differs, match it by what it asks for: an address, a key, a type.

## Step two, Connections: the model and its Context Length

Labels here, and in Connection Settings inside a chat, also follow the [February 2026 walkthrough](https://rpwithai.com/use-lorebary-on-wyvernchat/).

1. Open **Connections**, then **+ New Connection**.
2. **Name**: the model id you are about to choose, so a later connection for another model is easy to tell apart. **Visibility**: Private.
3. Switch **Force Chat Completion** on. Leave **Tool Call Enabled** off, **Inline Image Support** off and **Use Browser Request** off.
4. In **Model Configuration**, choose the **Provider** you created a moment ago.
5. **Model**: press **Fetch Models**. Wyvern calls `GET /v1/models` with your key and gets back every id the [Models page](https://unrestrictable.ai/models) lists; choose one, `uncensored-ai-roleplay` for instance. When the fetch fails, type the id by hand, spelled as the Models page spells it. A display name like "Uncensored AI Roleplay" is not an id and earns a 404 `model_not_found`.
6. **Context Length (tokens)**: take the figure for your model from the table in the price section. This number decides how much history goes out with each message, which is most of the bill, so a round maximum is the wrong choice.
7. Skip the **Instruct Templates** tab. Under **Parameters**, choose the **Stable** preset. No jailbreak text is needed: the builds listed here stay in character on their own. Press **Create**.

Last, inside a chat: the cog on the right opens **Connection Settings**. Turn **Free Queue** off and choose your connection under **Mine**. From the next message on, the model you named replies. A second model later means another connection under the same provider; address and key stay put, and one key reaches every id.

## Models worth putting in the field

With your own connection the request reaches the model as you wrote it: no prefill, no jailbreak attached by anyone in between. Four candidates, at the live catalog price:

| Model | Build | Input $/M | Output $/M |
| --- | --- | --- | --- |
| [Uncensored AI Roleplay](https://unrestrictable.ai/models/uncensored-ai-roleplay) uncensored-ai-roleplay | Roleplay | $0.60 | $2.34 |
| [GLM-5.3 Flash Uncensored](https://unrestrictable.ai/models/glm-5.3-flash-uncensored) glm-5.3-flash-uncensored | Uncensored | $0.1053 | $0.3685 |
| [GLM-5.3 Uncensored](https://unrestrictable.ai/models/glm-5.3-uncensored) glm-5.3-uncensored | Uncensored | $0.282 | $0.846 |
| [DeepSeek V4 Pro Uncensored](https://unrestrictable.ai/models/deepseek-v4-pro-uncensored) deepseek-v4-pro-uncensored | Uncensored | $0.97 | $1.94 |

Prices per million tokens, read from the live catalog on 2026-11-16.

- `uncensored-ai-roleplay`, [Uncensored AI Roleplay](https://unrestrictable.ai/models/uncensored-ai-roleplay): trained for roleplay, with a window of 128,000 tokens. The id we reach for first with a new character.
- `glm-5.3-flash-uncensored`, [GLM-5.3 Flash Uncensored](https://unrestrictable.ai/models/glm-5.3-flash-uncensored): the budget pick when a chat runs for hours.
- `glm-5.3-uncensored`, [GLM-5.3 Uncensored](https://unrestrictable.ai/models/glm-5.3-uncensored): the full GLM, for long scenes where the flash build feels thin.
- `deepseek-v4-pro-uncensored`, [DeepSeek V4 Pro Uncensored](https://unrestrictable.ai/models/deepseek-v4-pro-uncensored): the largest here, for a plot that asks for it.

Going smaller instead, the 24B roleplay fine-tune `cydonia-24b-v4.3` is at $0.60 per million output tokens right now. Every id, with abliterated and roleplay builds marked, is on the [Models page](https://unrestrictable.ai/models); the [uncensored hub](https://unrestrictable.ai/uncensored) says what the two kinds are. To try a model before you wire it up, the [uncensored AI roleplay](https://unrestrictable.ai/uncensored-ai-roleplay) page runs the same catalog in the browser.

A fixed prompt goes to every id once a day, and the reply is matched against what that model is known to say. The 2026-11-16 run covered 169 of the models this site lists; 147 replied as the named model. Each of the four above carries its own verdict on its model page. How the check works, and today's results, are at [/proof](https://unrestrictable.ai/proof).

## The price of a hundred messages, and the Context Length figure

What you type is a small part of a Wyvern message. The character card, your persona, every lorebook entry that matches and whatever history fits inside the Context Length are sent on every turn. The table prices one message at 4,000 tokens in and 300 out, then the same message deep into a long chat at 16,000 tokens in. Per-million prices are the live catalog's on 2026-11-16; the sum is tokens multiplied by price, with nothing added.

| Model | Per message | Per 100 messages | Per 100 at 16k context | Context Length value |
| --- | --- | --- | --- | --- |
| `uncensored-ai-roleplay` | $0.0031 | $0.3102 | $1.0302 | 65536 |
| `glm-5.3-flash-uncensored` | $0.0005 | $0.0532 | $0.1795 | 32768 |
| `glm-5.3-uncensored` | $0.0014 | $0.1382 | $0.4766 | 32768 |
| `deepseek-v4-pro-uncensored` | $0.0045 | $0.4462 | $1.6102 | 32768 |

The Context Length value is the catalog's window minus 16K, capped at 64K, or 32K when the developer publishes no window. Stay under the window: a prompt that overruns it is rejected rather than cut.

In practice the $10 minimum top-up covers roughly 3,223 messages of that size on `uncensored-ai-roleplay` at the current price. Credit stays valid for 12 months, so a quiet month costs you nothing. Per-token prices for every model, and the bonus tiers, are on the [pricing page](https://unrestrictable.ai/pricing).

## Confirm it from a terminal

Inside Wyvern, **Fetch Models** is the test: a populated list proves the address and the key. A terminal gives you the full response to read. The first command lists what your key can call; the second posts one message shaped like Wyvern's. Swap in your key and change nothing else.

```bash
curl https://api.unrestrictable.ai/v1/models \
  -H "Authorization: Bearer sk-unr-live-…"
```

```bash
curl -i https://api.unrestrictable.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-unr-live-…" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "uncensored-ai-roleplay",
    "max_tokens": 300,
    "temperature": 0.9,
    "messages": [
      {"role": "system", "content": "You are Ilse, the night warden of a border fort. Stay in character. Reply in two short sentences."},
      {"role": "user", "content": "Riders on the east road. Do we light the beacon?"}
    ]
  }'
```

Success is a 200 in the [chat completions format](https://platform.openai.com/docs/api-reference/chat). Its `usage` block is the basis of the charge; Wyvern's character, persona and history count in `prompt_tokens` exactly like the system message here. Each response carries two headers: `x-uplink-request-id`, to quote if you contact support, and `x-uplink-charge`, the amount settled for the call. Nothing is charged when a request fails before the model replies.

## What gets stored

Whoever runs the address in the Provider can read what you send there. Here is what this one records, as the [privacy page](https://unrestrictable.ai/privacy) puts it. A usage record for each request (its id, the model, token counts, the charge). A SHA-256 fingerprint of the request and of the response, never the text itself, so a signed receipt can be compared with the bill. Request bodies only if you switch replay on for a key, and then for 7 days only. Replay off means prompts and replies pass through unwritten. The key is kept as a salted hash and displayed once.

Two free precautions. Give the key a monthly spend limit in the console; a key that escapes in a screenshot then stops at the limit with a 402 `monthly_limit_reached` rather than emptying the balance. And use one key per app, Wyvern here and SillyTavern elsewhere, so revoking one leaves the other alone.

## When something breaks

| Symptom | Why | What to do |
| --- | --- | --- |
| 401, `invalid_api_key` | The key in the provider is mistyped, revoked, or belongs to another account. | Open the provider and paste a fresh copy of the key. |
| 404, `route_not_found` | The address lacks `/v1`. | Use `https://api.unrestrictable.ai/v1`; a second 404 means `https://api.unrestrictable.ai/v1/chat/completions`. |
| 404, `model_not_found` | A display name or a misspelling sits in the Model field. | Fetch Models once more, or paste the id in from the Models page. |
| 402, `insufficient_balance` | The balance is too low for the request. | Add credit in the console; the response body includes the link. |
| 429, `upstream_rate_limited` | That model's route is saturated. | Wait out `retry-after` and resend, or use another connection for a minute. |
| Nothing in the reply | The answer was cut off inside a reasoning step. | Allow 1,000 or more response tokens on a reasoning model. |
| The console's Usage page shows no request | The chat is still on another connection. | Connection Settings: Free Queue off, yours under Mine. |

Below, the bodies behind those status codes. The `code` field is stable, so match on it. None of these is billed.

```
HTTP 401, the stored key is wrong
{"error":{"message":"The API key is not valid.","type":"authentication_error","param":null,"code":"invalid_api_key"}}

HTTP 404, the API URL does not reach the chat completions route
{"error":{"message":"No such route.","type":"not_found_error","param":null,"code":"route_not_found"}}

HTTP 404, the Model field names an id that is not on the list
{"error":{"message":"No such model: gpt-4.","type":"not_found_error","param":null,"code":"model_not_found"}}

HTTP 402, the balance is empty
{"error":{"message":"The prepaid balance does not cover this request. Top up at … to continue.","type":"billing_error","param":null,"code":"insufficient_balance"}}
```

## Questions

### Can WyvernChat connect to a custom API?

It can. Under AI Connections you create a Provider (API URL Type OpenAI, an address, a stored key) and a Connection that names a model on it. Any chat completions API works there, including this one.

### Which URL goes in the Wyvern Chat API field?

https://api.unrestrictable.ai/v1, ending in /v1. Releases that post to the address exactly as typed, rather than adding the chat completions path, need https://api.unrestrictable.ai/v1/chat/completions.

### Does WyvernChat cost money?

Wyvern itself has a Free Queue and paid plans from $10 a month, per its plans page read on 2026-11-02. With your own key, Wyvern charges nothing and the model is billed per token from your balance.

### Fetch Models comes back empty in WyvernChat. Why?

Three usual causes: the address has no /v1, the key is mistyped, or No Key Required is switched on. Correct the provider, save it and fetch once more. Typing the id by hand also works. See the Models page.

### What model should I pick for roleplay in WyvernChat?

A roleplay or uncensored build, which holds character without any jailbreak. uncensored-ai-roleplay is the roleplay-trained default; the leaderboard orders the others by the daily check and by price. See the uncensored LLM leaderboard.

### Does one key work in both WyvernChat and SillyTavern?

Yes: all keys on the account draw from one balance. Still, give each app its own key: you can revoke one alone, and the Usage page lists them separately. See the SillyTavern setup.

Every model above is on the [uncensored model list](https://unrestrictable.ai/models) with its live price per million tokens. [Per-token pricing](https://unrestrictable.ai/pricing) sets the whole catalog side by side, and the [family directory](https://unrestrictable.ai/uncensored) groups the builds by base model.

[Get a key](https://app.unrestrictable.ai/console) Prepaid from $10, no subscription.

## New models and price changes, once a month.

New uncensored builds with their prices, price moves, and what the daily model check found.
