Intercept LLM API calls at the process level. No proxy. No MITM. No infra changes.
⚠️ This is a public bridge repository. The full source code, issue tracker, and development activity live in the privatenarko4u/vx2-gate-corerepository. This repo serves as the public-facing entry point.
LLM API calls are a blind spot. Every openai.ChatCompletion.create() and every anthropic.messages.create() is an unguarded tunnel out of your application.
- No visibility — who is calling what model, with what data?
- No control — can you stop a prompt leaking PII before it leaves your process?
- No security — prompt injection, secret exfiltration, data loss — all invisible.
- No governance — no audit trail, no rate limits per user, no content policies.
Your LLM integrations are running without guardrails. VX2 Gate fixes that with one line of code.
VX2 Gate SDK is a Python library that monkey-patches the official OpenAI and Anthropic SDKs at import time to insert a security, control, and observability layer — all inside your existing process.
pip install vx2-gateimport vx2_gate
vx2_gate.install()That's it. Every LLM call is now intercepted, inspected, rate-limited, and logged — with zero infrastructure changes.
| Feature | What it does |
|---|---|
| Monkey-patch intercept | Hooks openai and anthropic client methods at the Python level — no proxy, no network config |
| Content filtering | Strips PII (emails, SSNs, API keys), secrets, and blocks prompt injection attempts before they reach the API |
| Rate limiting | Per-model and per-user throttling — prevent runaway costs and abuse |
| Audit logging | Full request/response logs with timestamps, user IDs, model names, token counts |
| Transform hooks | Intercept and mutate requests/responses in-flight — redact, enrich, or block |
| Zero infra change | No proxy to deploy, no DNS to change, no certificates to manage — it runs in your process |
pip install vx2-gateCreate a config file (vx2_config.yaml):
gate:
rate_limit:
openai/gpt-4: 10/minute
anthropic/claude-3-opus: 5/minute
filter:
block_pii: true
block_prompt_injection: true
audit_log: ./logs/vx2-audit.jsonlInstrument your app:
import vx2_gate
vx2_gate.install(config="vx2_config.yaml")
# Your existing OpenAI/Anthropic code — unchanged
import openai
response = openai.ChatCompletion.create(
model="gpt-4",
messages=[{"role": "user", "content": "Hello"}]
)Every call is now filtered, rate-limited, and logged. No proxy. No infra.
- Enterprise Security — prevent data leakage through LLM APIs across your entire Python application
- Compliance Auditing — log every prompt and response for SOC 2, HIPAA, GDPR audits
- Developer Debugging — see exactly what your agents are sending to and receiving from LLMs
- Rate Limiting — control costs by limiting per-user and per-model call volume
- Content Governance — enforce corporate policies on what can and cannot be sent to LLMs
Traditional LLM gateways are reverse proxies that sit between your app and the API provider. They require:
- Network configuration and DNS changes
- TLS certificate management
- Dedicated infrastructure to deploy and maintain
- Latency from an extra network hop
VX2 Gate runs in-process. There is no network hop, no deployment, no config changes. It intercepts at the Python SDK level — the lowest possible point before the call leaves your process. This means:
- Zero added latency
- Zero infrastructure
- Zero network config
- Works in serverless, containers, Jupyter notebooks — anywhere Python runs
The full source code for VX2 Gate is maintained in the private narko4u/vx2-gate-core repository. This public repo provides documentation, issue tracking, and community engagement.
| Item | Price |
|---|---|
| SDK License (perpetual, per org) | $999 |
| Monthly support & updates | $199/mo |
Includes: SDK license key, configuration templates for 10+ common security policies, email support, and regular SDK updates.
Instant delivery. License key sent via email.
Built by Empire Labs Pty Ltd (ACN 693 862 145)