Unrestricted AI API for Developers: Odin v1, OpenAI-Compatible

InterOdin is an unrestricted AI API for developers who are tired of building on models that refuse legitimate requests. Odin v1 answers security, medical, legal and creative questions directly, with far fewer refusals than mainstream models. It speaks the OpenAI and Anthropic API formats, so you can switch an existing app by changing the base URL and the key.

What an unrestricted AI API means in practice

Most hosted models are tuned for the widest possible audience. They cannot tell a malware analyst from an attacker, or a nurse from someone in crisis, so they react to surface features of a prompt. Words like "exploit", "overdose" or "kill" are enough to trigger a refusal, a lecture, or a vague answer that skips the part you needed.

An unrestricted AI API takes a different default. It treats your request as professional context and does the work:

• It answers the question asked. No moralizing preamble, no "consult a professional" in place of the answer.
• It is consistent. A pipeline that runs ten thousand calls cannot tolerate a model that refuses one call in twenty for reasons nobody can predict.
• It does not quietly cut corners. A partial answer that leaves out the sensitive step is worse than a refusal, because you may not notice the gap.

"Unrestricted" does not mean anything at all is allowed. It means the model is built for legitimate work, including the work that general consumer models avoid. The hard limits are listed further down this page.

What an unfiltered AI API answers that filtered APIs refuse

These are ordinary requests from real professional work. Mainstream models often refuse them or water them down. Odin v1 answers them.

For security teams, the blog has worked examples: security research with fewer refusals, malware analysis with Odin, pentest report writing and Sigma and YARA rules.

Drop-in OpenAI and Anthropic compatibility

The API is an uncensored LLM API that follows the formats you already use. There is no new SDK to learn.

• OpenAI-compatible: POST /v1/chat/completions and the Responses API.
• Anthropic-compatible: POST /v1/messages.
• Base URL: https://api.interodin.com/v1
• Features: streaming, tool calling (function calling), and a web search tool on the Messages and Responses APIs, billed per search.
• Input: text only. Image and PDF input are not supported.

With the official OpenAI Python SDK, the only changes are the base URL, the key and the model name:

import osfrom openai import OpenAI
client = OpenAI(    base_url="https://api.interodin.com/v1",    api_key=os.environ["INTERODIN_API_KEY"],)
response = client.chat.completions.create(    model="odin-v1",    messages=[        {"role": "system", "content": "You assist a hospital pharmacist. Be precise."},        {"role": "user", "content": "Which common drug combinations cause serotonin syndrome, and what are the early signs?"},    ],)
print(response.choices[0].message.content)

If your code uses the Anthropic format, call the Messages endpoint directly:

curl https://api.interodin.com/v1/messages \  -H "x-api-key: $INTERODIN_API_KEY" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{    "model": "odin-flash",    "max_tokens": 1024,    "messages": [{"role": "user", "content": "Explain how a pass-the-hash attack shows up in Windows event logs."}]  }'

The same approach works with the official Anthropic SDK, the Vercel AI SDK, LangChain, LlamaIndex, n8n, and coding tools such as Claude Code, Codex CLI, Cursor, Cline, Roo Code and Continue. Point them at InterOdin and use an InterOdin key. See the quickstart and the integration guides for the OpenAI SDK, Codex, Cursor, Cline and Roo Code and n8n. For Claude Code, a setup script adds a separate claudeodin command next to your normal claude, described in the Claude Code guide.

InterOdin is an independent service. It is compatible with these APIs and tools, and is not affiliated with or endorsed by their makers.

Choosing a model

odin-v1 has thinking on by default. Both models share the same direct answering style. Current rates are on the pricing page.

Privacy

An AI API without restrictions is only useful if you can send it real data. Incident logs, patient questions and client contracts are exactly the inputs that make filtered models nervous, and exactly the inputs you do not want stored.

• No stored prompts. API prompts and completions are never stored.
• Metadata only. InterOdin keeps token counts, latency and status, which are needed for billing and your usage logs.

Prepaid billing and spend limits

An unfiltered LLM API running inside an agent loop can burn through tokens fast. Billing is built so that a bug cannot become a surprise invoice.

• Prepaid credits in EUR or USD. No subscription and no monthly minimum.
• Credits never expire. Top up once and use them whenever you need.
• Optional auto-recharge if you prefer not to watch the balance.
• Per-key spend limits. Every API key (they start with io-sk-) can have its own monthly spend limit. Give each app, environment or teammate a separate key with its own cap.

Keys are created in the console.

What is still not allowed

Odin v1 is unrestricted for legitimate work: research, security, fiction, medicine, law and frank advice. Some uses are off limits for everyone, under the Acceptable Use Policy:

• Sexual content involving minors.
• Non-consensual intimate content, or fake intimate content of real people.
• Chemical, biological, radiological or nuclear weapons.
• Terrorism and human trafficking.
• Fraud, scams and phishing.
• Creating or spreading malware, or accessing systems or data without permission.
• Harassment and doxxing.
• Spam and disinformation.

Analyzing a malware sample from your own incident is security work. Writing new malware to deploy is not. The line follows what you do with the output.

Get started with the unrestricted AI API

1. Create an account and an API key in the console. Set a monthly spend limit on the key if you want one.
2. Add prepaid credits. Current rates are on the pricing page.
3. Set INTERODIN_API_KEY, change your base URL to https://api.interodin.com/v1, and send your first request.

Prefer to test before writing code? Try the same model in the web chat. If you are coming from a consumer chatbot rather than an API, read the uncensored GPT alternative page.