How to Integrate the Uncensored AI API

Why Choose an Uncensored AI API?

Standard LLMs often apply content filters that block creative writing, controversial opinions, or adult themes, even when those topics are perfectly lawful. An uncensored AI API removes these artificial boundaries, giving you raw model outputs. This is critical for developers building creative tools, roleplay engines, or research models where nuance and lack of refusal matter more than brand-safe compliance.

Unlike model aggregators that route requests through multiple vendors, a dedicated uncensored API offers consistency. You get one specific model tuned to answer freely. There is no hidden filtering layer applied by a third-party aggregator. This direct line ensures that if the model can generate the content, it will. The trade-off is that you manage your own content moderation if needed, but you gain full control over the output style and depth.

Setting Up Your Environment

Integration begins with configuring your development environment to point to the correct base URL. Since this API is OpenAI-compatible, you likely already have the necessary SDKs installed. You do not need new libraries; you only need to update the endpoint configuration.

• Base URL: https://api.uncensoredai.top/v1
• Model ID: "uncensored"
• Protocol: HTTPS

This setup works with the official OpenAI SDKs for Python, Node.js, and other languages. Simply update the base_url parameter in your client configuration. The API serves a single model, so you do not need to manage model routing or versioning. This simplicity reduces integration time from hours to minutes. You can start sending requests immediately after obtaining your API key.

Authentication and Headers

Authentication is handled via a single API key passed in the Authorization header. This is the standard method used by OpenAI and compatible services. You do not need complex OAuth flows or temporary tokens for basic usage.

• Header Name: Authorization
• Header Value: Bearer YOUR_API_KEY

Your API key is generated when you sign up via Google or email. It is shown immediately and should be stored securely. Each account is limited to one active key at a time; generating a new key invalidates the previous one. This design keeps authentication simple and secure. There are no per-request authentication tokens or complex signing mechanisms to manage.

Sending Your First Completion Request

Sending a request follows the standard POST /v1/chat/completions endpoint. You provide the model name, messages, and optional parameters. The API returns a text response. This is a pure text-in, text-out interface.

• Endpoint: POST /v1/chat/completions
• Model: "uncensored"
• Input: JSON body with messages array

The model has a context window of 64,000 tokens, allowing for long conversations or large document processing. The maximum output per request is 16,000 tokens, or 2,048 if you do not specify max_tokens. This capacity is sufficient for most creative writing and analytical tasks. You can adjust parameters like temperature and top_p to control creativity.

Handling Streaming Responses

For better user experience, especially in chat interfaces, use Server-Sent Events (SSE) to stream responses. This allows users to see tokens as they are generated, reducing perceived latency. The API supports streaming natively.

• Stream Parameter: Set stream: true in your request.
• Response Type: Server-Sent Events
• Token Usage: Sent in the last chunk

Streaming is essential for applications where real-time feedback is expected. The API sends chunks of text as they are generated. The final chunk includes token usage statistics, allowing you to track costs accurately. This feature works seamlessly with standard OpenAI-compatible streaming clients. You do not need custom parsing logic for standard SSE formats.

Advanced Features: JSON Mode and Tools

The API supports advanced features like JSON mode and function calling, making it suitable for structured data extraction and agent workflows. You can enforce strict JSON output using the response_format parameter.

• JSON Mode: Set response_format to {"type": "json_object"}
• Function Calling: Support for tools and tool_choice
• Parameters: temperature, top_p, stop, seed, presence_penalty, frequency_penalty

These features allow the uncensored model to integrate into complex pipelines. You can extract structured data from unstructured text or control model behavior with custom tools. The model handles these requests reliably, maintaining its uncensored nature even when constrained to JSON format. This flexibility makes it a powerful tool for developers building custom AI applications.

Monitoring Usage and Limits

Understanding rate limits and usage tracking is crucial for stable integration. The API enforces specific limits to ensure fair usage across all clients.

• Rate Limit: 300 requests per minute per key
• Concurrent Requests: 8 simultaneous requests per key
• Max Body Size: 8 MB per request

These limits are generous for most development and production use cases. If you exceed the rate limit, you will receive a standard error response. Monitoring your token usage helps manage costs, as pricing is based on actual token consumption. Errors and refusals do not consume credit, so you only pay for successful completions. This pay-as-you-go model ensures you never pay for wasted requests.

Managing Credits via Crypto

Pricing is straightforward: $0.25 per 1M input tokens and $1.00 per 1M output tokens. There are no monthly subscriptions or hidden fees. You top up your account with crypto, and credit is charged based on real token usage.

• Payment Methods: USDT (TRC20) or USDC (Base)
• Top-up Range: $10 to $500
• Bonus: +5% for $50+, +10% for $100+

Credit never expires, so you can top up when convenient. Errors are free, meaning you only pay for successful completions. This model is ideal for developers who want predictable costs without the commitment of a subscription. You can start with a $0.50 trial credit, valid for 7 days, with no card needed. This low barrier to entry makes testing and integration easy.