Security-specialized models in our app or through one API, with zero data retention. Subscribe, or pay per token.
Works with your languages and tools
A general-purpose model knows a little about everything. Ours are trained on logs, incident reports and vulnerabilities. They know what they’re looking for.
Pick by task. Same API, same response format, same key.
The lightest model. It classifies an alert, an email or a log line in milliseconds, fast enough to filter a live stream.
The everyday model. It reads entire logs, summarizes an incident and writes your SIEM queries from a plain-English question.
The most capable. It reasons over a full attack chain, reviews code line by line and proposes a well-argued fix.
Keep your code. Point base_url at unbreach, pick a model, done. Read the API docs
from openai import OpenAI
import os
client = OpenAI(
base_url="https://api.unbreach.io/v1",
api_key=os.environ["UNBREACH_API_KEY"],
)
resp = client.chat.completions.create(
model="rampart-1",
messages=[{
"role": "user",
"content": "Analyze this log: "
"Failed password for root from 185.220.101.4 (x312)",
}],
)
print(resp.choices[0].message.content)import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.unbreach.io/v1",
apiKey: process.env.UNBREACH_API_KEY,
});
const resp = await client.chat.completions.create({
model: "rampart-1",
messages: [{
role: "user",
content: "Analyze this log: " +
"Failed password for root from 185.220.101.4 (x312)",
}],
});
console.log(resp.choices[0].message.content);curl https://api.unbreach.io/v1/chat/completions \
-H "Authorization: Bearer $UNBREACH_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "rampart-1",
"messages": [{
"role": "user",
"content": "Analyze this log: Failed password for root from 185.220.101.4 (x312)"
}]
}'0days of data retention
Your prompts and our answers are processed in memory, then discarded. Nothing you send is written to disk, logged or used for training, in the app or through the API.
How we handle dataDrop the false positives and keep only what deserves an analyst’s attention.
Ask a question in plain English, get the SIEM query and the answer.
query = f"SELECT * FROM users WHERE id = {user_id}"
SQL injection. Use a parameterized query.
Injections, exposed secrets, vulnerable dependencies: caught before production.
Headers, links and tone analyzed in a single request.
A raw timeline becomes a clear report, ready for leadership.
A monthly plan for the app, or pure pay-per-token through the API. Zero data retention on both.
For people who check their security every day.
$20 / month
Billed monthly
Get PlusIncludes:
For heavy users, freelancers and consultants.
From $100 / month
Billed monthly
Get ProEverything in Plus, plus:
Usage limits apply. Prices in USD, excluding applicable taxes. Plans renew automatically until you cancel.
No subscription, no minimum. Billed monthly on what you actually use.
Public per-token pricing, no commitment.
from $0.15Automatic discount on monthly spend above $1,000.
-20%Dedicated instance, sovereign or on-premise deployment, contractual SLA.
Talk to the teamModel
$105 / month
About 37 million words processed.
Volume discount applied.
Assumes 75% input tokens and 25% output tokens.
A token is a piece of a word. In English, 1,000 tokens are roughly 750 words. Input tokens (your request) and output tokens (the response) are counted separately.
Each plan comes with a usage allowance that refills every hour. How far it goes depends on the model and the length of your conversations: bastion-1 uses it faster than sentry-1. Pro gives you 5x or 20x more usage than Plus.
You can wait for the next refill, switch to a lighter model, or add pay-as-you-go credits at API prices. We never charge you more without asking first.
Yes, at any time from your account. You keep access until the end of the month you already paid for.
No. unbreach is zero data retention by default, for the app and the API. Prompts, responses and uploaded files are processed in memory and discarded once the answer is sent. We only keep token counts and billing records, and nothing you send is ever used for training.
Yes. The API follows the OpenAI SDK format: change the base URL and the model name, and the rest of your code works as is.
In data centers located in the European Union. The Enterprise plan adds a dedicated instance or a deployment on your own servers.
Plans are a fixed price, billed monthly or yearly and renewed until you cancel. API usage is billed monthly on what you actually use, or from prepaid credits with a spending cap per key.
Yes. They grow automatically with your usage history. For a one-off spike, write to us at [email protected].
Tell us who you are. We review every request within one business day, then you create keys and make your first request in minutes.