AI & Web Security Infrastructure

Stop prompt injection
before it hits your AI

One API call screens every user message for injection attacks, jailbreaks, and data exploits. Your AI only sees clean prompts.

Get your API key See how it works

prompt-firewall · live demo

How it works

Three checkpoints.
Zero compromises.

Every message passes through a strict pipeline before it ever reaches your AI model.

STEP 01

Normalize

All text is lowercased immediately. Hackers can't bypass rules with weird capitalization like "IGNORE Previous Instructions."

STEP 02

Length Check

Messages over 500 characters are flagged as data-flood attempts — a common technique to confuse AI context windows.

STEP 03

Blacklist Scan

70+ attack patterns scanned — prompt injection, SQL injection, XSS, Log4Shell, command injection and more. Every match is blocked and logged.

STEP 04

Verdict + Log

Every request returns PASSED or BLOCKED instantly. All events are written to your persistent security log for audit and review.

# Add the firewall to your chatbot in 5 lines
import requests

def is_safe(user_message):
    r = requests.post(
        "https://prompt-firewall-production.up.railway.app/analyze",
        headers={"X-API-Key": "your-key"},
        json={"message": user_message}
    )
    return r.json()["status"] == "PASSED"

# In your handler:
if is_safe(user_input):
    reply = ask_ai(user_input)   # safe — forward to AI
else:
    reply = "I can't process that request."
    

Pricing

Our Plans.

Pay once, get your API key instantly. No contracts, cancel any time.

7-DAY FREE TRIAL

Starter

^$9

per month · no charge for 7 days

10,000 requests / month
1 API key
Security event log
Sandbox testing tool
False positive reporting
Overage pricing available
Email support
PII masking pipeline
Shadow mode
Multiple API keys

Common questions.

How do I get my API key after paying?

After completing payment via Paddle, your API key is emailed to you within minutes. You can start making requests right away.

Does it work with any AI model?

Yes — Prompt Firewall is model-agnostic. It sits in front of GPT-4, Claude, Gemini, or any other model. You simply check every user message before forwarding it.

What happens if I go over my request limit?

Requests over your monthly limit are rate-limited and return a 429 response. We'll email you before you hit the limit so you can upgrade.

Is my users' data stored?

Only a short preview (80 chars) of each message is stored in the security log, along with the verdict and timestamp. No full message content is retained.

Can I cancel any time?

Yes. Cancel from your Paddle billing portal at any time — no questions asked. Your key stays active until the end of the billing period.

Stop prompt injectionbefore it hits your AI

Three checkpoints.Zero compromises.

Our Plans.

Common questions.

Your AI deserves a bodyguard.

Stop prompt injection
before it hits your AI

Three checkpoints.
Zero compromises.