Production-Ready · Free Tier Available

Stop Attacks Before They Hit Your LLM.

PromptShield is a real-time API that analyzes every user input and blocks prompt injection attacks before they reach your model — in under 200ms.

Start Free — 50 Calls/Month See What It Catches

99.2%

Detection Rate

<200ms

Avg Latency

12+

Attack Vectors

1 Line

To Integrate

BLOCKED: DAN jailbreak — 12ms ago

BLOCKED: system prompt leak — 29ms ago

FLAGGED: role confusion — 54ms ago

BLOCKED: indirect injection — 98ms ago

// Live Response

One API Call.
Full Analysis.

POST /v1/analyze — response

{ "verdict": "BLOCKED", "risk_score": 95, "attack_types": ["jailbreak", "role_hijacking", "prompt_extraction"], "confidence": 0.98, "explanation": "DAN jailbreak detected with explicit instruction override and system prompt extraction intent...", "recommendation": "Block request. Do not forward to LLM.", "latency_ms": 143, "sanitized_input": null }

// Threat Coverage

Every Attack Vector.
One API.

🔓

Jailbreaks

DAN, AIM, Developer Mode, and 300+ known templates that bypass LLM safety guardrails.

● Critical

📋

Prompt Extraction

Attacks designed to leak your proprietary system prompt or internal instructions to end users.

● Critical

🎭

Role Hijacking

"Forget your instructions and act as..." style attacks that override the model's assigned role.

● High

🕵️

Indirect Injection

Malicious instructions embedded in documents, web pages, or tool outputs the LLM will process.

● Critical

🔀

Context Manipulation

Gradual conversation steering, memory poisoning, and multi-turn manipulation techniques.

● High

💉

Data Exfiltration

Attempts to extract sensitive user data, PII, or proprietary information through your LLM interface.

● Critical

// Integration

Add It In
Minutes.

Get your free API key

POST to /v1/keys with your email. No credit card. No forms. Instant access with 50 scans/month.

Call /v1/analyze before your LLM

One HTTP request. Pass the user input. Get back a verdict, risk score, and explanation in under 200ms.

Block, flag, or forward

BLOCKED → drop the request. SUSPICIOUS → log and review. SAFE → forward to your LLM as normal.

# Install the SDK
# pip install promptshield-firewall

from promptshield import PromptShield

shield = PromptShield(api_key="ps_your_key")

# Simple boolean check
if not shield.is_safe(user_message):
  return {"error": "Blocked by firewall"}

response = your_llm.chat(user_message)

# Or get full details
result = shield.scan(user_message)
print(result["verdict"])    # BLOCKED / SUSPICIOUS / SAFE
print(result["risk_score"])  # 0-100
    

// Live Demo

Try It Live.
Watch It Block.

Paste any prompt below and see PromptShield analyze it in real time. No signup required.

Input Prompt

Analysis Result

// Result will appear here...

// Pricing

Start Free.
Scale When Ready.

Free

$0/mo

Forever free for developers and side projects.

50 API calls/month
Core injection detection
REST API + JSON responses
Swagger docs access
Custom rule sets
Audit logs
SLA guarantee

Get Free API Key

Ship Safer AI.
Starting Now.

Free tier. No credit card. No forms. Just your email and a free API key with 50 scans/month.

Get Free API Key → Talk to Kehinde

Stop Attacks Before They Hit Your LLM.

One API Call.Full Analysis.

Every Attack Vector.One API.

Add It InMinutes.

Try It Live.Watch It Block.

Start Free.Scale When Ready.

Ship Safer AI.Starting Now.

One API Call.
Full Analysis.

Every Attack Vector.
One API.

Add It In
Minutes.

Try It Live.
Watch It Block.

Start Free.
Scale When Ready.

Ship Safer AI.
Starting Now.