PromptZero v2.3.0 — Zero trace. Full answer.

PromptZero v2.3.0

Zero trace. Full answer. — Cero rastro. Respuesta real.

PromptZero is a local, transparent proxy for the Claude API that detects and replaces sensitive data in your prompts before they leave your environment — then restores real values in the response.

Built by OpenBash.com — from pentesters, to pentesters.

---

What's included

Core proxy

• Transparent drop-in replacement for https://api.anthropic.com
• Dual detection layer: NLP (Presidio + spaCy) + Regex
• Session-scoped bidirectional mapping table
• Streaming (SSE) support
• Graceful degradation to regex-only if NLP not installed

What gets protected

Type	Real → Synthetic
Person names	John Smith → Alice Harrington
Organizations	Acme Corp S.A. → Globex Industries
IPv4 (pentesting)	192.168.1.45 → 127.0.0.1
Hostnames	db.prod.corp.com → localhost.localdomain.1
Emails	john@corp.com → user001@fakecorp.local
National ID / DNI	28.456.123 → FAKE-ID-000001
Phones, SSN, IBAN, Credit cards, API keys	✓

Examples

• document_summary/ — Summarize PDF, DOCX, TXT with full PII protection
• pentest_report/ — Generate technical/executive/remediation reports from findings JSON

---

Quick Start

git clone https://github.com/openbashok/promptzero
cd promptzero
./setup.sh
cp .env.example .env  # add your ANTHROPIC_API_KEY
python main.py

One line change in your code:

client = anthropic.Anthropic(
    api_key="your-key",
    base_url="http://localhost:8000",  # ← only change
)

---

Why PromptZero?

Every prompt you send to an AI contains real IPs, real names, real credentials, real infrastructure details. That data goes to a third-party server — every time. PromptZero fixes this.

The AI sees fiction. You get facts.

---

Roadmap

• Persistent sessions (SQLite / Redis)
• Multilingual NLP (ES, PT, FR, DE)
• log_analyzer and code_reviewer examples
• Docker image
• CLI wrapper

---

Made with ♥ by the OpenBash community — from pentesters, to pentesters.