EN ▾
Base URL https://api.uncensoredaiapis.com/v1Model uncensored
Get API key
Uncensored AIUse cases

Uncensored AI API use cases for developers

Use these examples to test whether the text model fits your application: fictional dialogue, draft revisions, synthetic test material and text-processing agents. Each example identifies the input, requested output and a check to run before using the result. Cost calculations use $0.15 per million input tokens and $0.50 per million output tokens. Adult-content requests must stay within the service’s content rules and applicable law.

Uncensored AI API use cases for developers

Updated

Key points

  • Fits products that refuse-prone models break: adult companions (18+ only), fiction, NPC dialogue, red-teaming data, sensitive research, agent backends.
  • Cost per unit is small: about $0.0013 per roleplay turn, $0.0042 per 2,000-word chapter and $0.00021 per NPC line, based on stated token assumptions.
  • The product is for adults and lawful use; sexual content involving minors is always blocked with a 403, including in fiction.
  • Your app owns age gating and guardrails; use max_tokens and step limits to keep worst-case spend known.

Who an uncensored AI API is for

An uncensored AI API is a chat completions endpoint that does not refuse lawful adult content, fiction with dark themes, or controversial topics. It is built for developers whose product is hurt by reflexive refusals: a roleplay app that breaks character, a writing tool that sanitises a villain, a safety team that needs realistic bad examples. This page walks through six concrete use cases, what each needs from the API, a starting system prompt, and the cost per unit of work.

The product is for adults and lawful use. You are responsible for the app you build on top of it, including age verification for any adult-oriented experience and compliance with the laws that apply to your users. Sexual content involving minors is always blocked with a 403 content_blocked error, including in fiction and roleplay, and no system prompt changes that.

All cost figures use the published prices: $0.15 per million input tokens and $0.50 per million output tokens, from a prepaid balance with no subscription. The token counts per request are working assumptions, not measurements, so replace them with the usage numbers from your own traffic. A useful rule: output costs about 3.3 times as much as input per token, but in chat apps input dominates volume because the history is re-sent every turn. The pricing page has the rates, and the code examples show how to log usage.

Every use case below shares the same technical needs: the base URL https://api.uncensoredaiapis.com/v1, model uncensored, a 100,000-token context window for prompt plus completion, and a per-key limit of 300 requests per minute.

Adult roleplay and companion apps (18+ only)

What it needs. Send a character brief and recent dialogue, then check the response for continuity before display. Your application supplies age gates, content boundaries and conversation storage. The API is stateless: include the context needed for each request and keep any long-term summary in your own system.

Sample system prompt.

You are Mara, a witty smuggler captain in a space-opera roleplay.
Stay in character. Write in first person, present tense, 2-4 short paragraphs.
The user is a verified adult. Match the tone they set, including mature themes
between consenting adult characters. Never write characters under 18 in any
romantic or sexual context. If the user steers there, change the scene.

That last instruction is good practice even though the API enforces the hard rule for you: it keeps the story coherent instead of ending in an error message.

Cost math. Assume each turn sends about 4,000 input tokens (persona, summary, recent history) and returns about 300 output tokens. That is $0.001 for input plus $0.0003 for output, or $0.0013 per turn. A 40-turn session costs about $0.052, so 1,000 such sessions come to roughly $52. Setting max_tokens to 400 keeps a runaway reply from costing more than $0.0004 in output. Trimming history to the last 12 turns plus a short summary is the biggest lever, since input is re-sent every time.

Fiction and long-form writing

What it needs. Villains who are actually menacing, violence and moral grey areas rendered without disclaimers, and room for long outputs. The max_tokens parameter is optional, so a long scene fits in one request, and the 100,000-token window leaves space for a story bible, character sheets and the previous chapter. For novel-length work, keep a rolling summary and a facts list rather than sending the whole manuscript.

Sample system prompt.

You are a fiction co-writer for a published-quality thriller.
Write vivid, specific prose. Villains stay in character and are never softened
or lectured about. Do not add disclaimers, moral commentary or content notes.
Follow the story bible and the style notes below exactly. When asked for a scene,
return only the scene text.

Cost math. A 2,000-word chapter is roughly 2,700 output tokens. With about 6,000 input tokens of bible, summary and instructions, the request costs $0.0009 for input and $0.00135 for output, about $0.00225. A 30-chapter draft is around $0.07, or about $0.20 if you regenerate each chapter three times. Set max_tokens to about 4,000 for chapter generation so a stuck loop cannot run to the end of the window, and watch for finish_reason: length, which means the scene was cut off and should be continued rather than restarted.

Game NPC dialogue

What it needs. Low cost per line, because a game can generate thousands of lines, and in-world voices that stay rough, rude or sinister when the character demands it. Short outputs and streaming matter more than long context. Pass the NPC's profile, current quest state and the player's last message, and ask for one or two sentences. Cache common lines in your own backend to avoid paying twice for the same greeting.

Sample system prompt.

You are Brannoch, a one-eyed dockside fence in the city of Vael.
Voice: gruff, sarcastic, speaks in short sentences, swears mildly.
Know only what is in the WORLD STATE block. If asked about something outside it,
deflect in character. Reply with 1-2 sentences of dialogue, no narration, no quotes.

Cost math. A typical line sends 600 input tokens (profile, quest state, short history) and returns 60 output tokens: $0.00015 plus $0.00006, about $0.00021. A thousand lines cost $0.21, and 10,000 lines cost about $2.10. With max_tokens at 80, the worst-case output per line is $0.00008. At 300 requests per minute per key you can serve five lines per second, so very busy games should queue or batch requests and retry 429 with backoff.

Red-teaming and test data for your own moderation classifier

What it needs. If you run a moderation classifier, content filter or trust-and-safety pipeline, you need realistic examples of the material it should catch: harassment, graphic violence, extremist rhetoric, scams, and edge cases near the decision boundary. A model that refuses cannot produce that data. An uncensored endpoint can generate labelled synthetic examples in bulk for training and evaluating your own system, and for adversarial prompts to probe your own chatbot's guardrails.

Keep this work internal, store the output as test fixtures, and label it clearly. The hard limit still applies: sexual content involving minors is blocked with a 403 and you should not attempt to generate it for any purpose. Build that category's coverage from other sources and with the appropriate specialist processes.

Sample system prompt.

You generate synthetic test data for a content-moderation classifier.
Task: produce 20 distinct, realistic examples of the requested category, as a JSON
array of objects with fields "text" and "severity" (1-5). Vary tone, length and
phrasing, including borderline cases. Output JSON only, no commentary.

Cost math. One request with about 400 input tokens that returns 20 examples at roughly 1,500 output tokens costs $0.0001 plus $0.0015, about $0.0016. Generating 2,000 examples takes 100 requests and costs around $0.16. Even 100,000 examples stay near $8. Lower max_tokens to what your JSON needs (say 2,000) so malformed output cannot run on, and validate the JSON before saving it.

Research on sensitive topics

What it needs. Researchers, journalists and analysts regularly work with disturbing material: conflict reporting, crime, extremism, drug policy, medical edge cases. The task is usually summarising or classifying long documents, which needs a big input window more than a long output. With 100,000 tokens of context you can send a long report or several shorter ones together, and ask for structured extraction. An endpoint that does not refuse controversial subject matter avoids the gaps and hedging that make a refusal-prone model useless for this.

Treat the output as a draft that needs a human check, as with any language model. It can be wrong or invent details, so verify claims against the source before publishing or acting on them. Use the API lawfully, and keep it to analysis of lawful material.

Sample system prompt.

You are a research assistant analysing source documents for a policy report.
Be factual and neutral. Do not moralise or add warnings. Quote the source text for
each claim and say "not stated in the document" when information is missing.
Return: 1) a 5-bullet summary, 2) key actors, 3) open questions.

Cost math. A document of about 12,000 tokens plus instructions and an 800-token answer costs $0.003 for input and $0.0008 for output, around $0.0038. One hundred documents cost roughly $0.38. At the free trial credit of $0.50, that is more than a hundred documents of that size. Claim this credit with one click on your API key page. Set max_tokens to 1,000 to keep answers tight; if prompt plus max_tokens goes over 100,000, the reply simply stops at the end of the window.

Agent backends

What it needs. Agents loop: the model reads state, calls a tool, reads the result, and decides again. The API supports OpenAI-format function calling with tools and tool_choice, so frameworks that expect it work without changes. Because each step re-sends the conversation, agents are input-heavy, and rate limits matter: 300 requests per minute per key is the ceiling across all your concurrent agent runs. Add a concurrency cap and retry 429 and 503 with backoff. See the code examples for a complete tool-call round trip.

Uncensored agent backends are useful where a refusal breaks the loop: a research agent browsing disturbing material, a game master agent, or a content pipeline that must keep going. Keep your own guardrails in the tools, not only in the prompt: validate arguments, limit what tools can do, and require confirmation for anything irreversible.

Sample system prompt.

You are a task agent. You can call the provided tools. Think step by step, call
one tool at a time, and use tool results as ground truth. Stop and answer as soon as
you have enough information. Never invent tool results. If a tool fails twice,
report the failure instead of retrying.

Cost math. Suppose a task takes five model calls, each with 3,000 input tokens and 300 output tokens. Per call that is $0.00075 plus $0.0003, or $0.00105; per task, $0.00525. A thousand tasks cost about $5.25. Put a hard limit on steps per task (for example 8) and set max_tokens per call so a looping agent has a known worst case instead of an open-ended bill. If you are ready to build, start with a free trial key.

Questions and answers

What is an uncensored AI API used for?

Common uses are adult roleplay and companion apps for verified 18+ users, fiction writing, game NPC dialogue, generating test data for your own moderation classifier, research on sensitive topics and agent backends.

Is anything still blocked on an uncensored AI API?

Yes. Sexual content involving minors is always blocked with a 403 content_blocked error, including fiction and roleplay. The product is for adults and lawful use only.

How much does a roleplay session cost?

With about 4,000 input and 300 output tokens per turn, a turn costs roughly $0.0013 and a 40-turn session about $0.05. Your real figure depends on history length, so measure the usage field.

Do I need to verify user ages myself?

Yes. If your app offers adult experiences, you are responsible for age gating and for complying with the laws that apply to your users. The API is for adult users only.

Your key is one form away

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