ai-devsec-gateway 1.6.0

Take back control. Intercept, audit, and route your AI traffic.

🛡️ AI DevSec Gateway (formerly AI Network Blocker)

Local controls for blocking, auditing, and routing AI developer traffic.

English | Español

---

📖 What is this?

AI DevSec Gateway is an open-source privacy and DevSecOps tool for developers adopting AI coding assistants. It provides local controls to block known AI endpoints, route API traffic to local inference servers, and audit active AI editor processes.

Originally created as a simple GUI to block AI endpoints, it is evolving into a Zero-Trust Gateway for safer AI-assisted development. The current release focuses on deterministic hosts-file blocking, a local HTTP router, safe-by-default CLI controls, and a security auditor that keeps API keys in memory only.

1. Block: A deterministic OS-level override via the hosts file that drops connections to 38+ known AI domains.
2. Route: A local HTTP proxy that can route compatible API clients to local LLMs such as Ollama, LM Studio, or vLLM.
3. Audit: Process-aware security checks that help identify active AI tools and data-leak risk signals.

---

✨ Features

Feature	Description
📊 Live Token Dashboard	Real-time token in/out and request counts with auto-refresh every 5s while the gateway runs. Fetch stats via /stats HTTP endpoint.
🔀 Transparent API Router	Seamlessly reroute Copilot/Cursor HTTP traffic to your own Local LLM inference servers.
🛡️ AI DevSec Auditor	Live process analysis with on-demand OpenAI recommendations. API keys are memory-only and never persisted.
🔄 DLP & Guardrails Toggle	Enable or disable Deep Packet Inspection and semantic guardrails per-session via gateway config flags.
💻 Native CLI Interface	Full headless control for CI/CD environments. Use ai-blocker --status or ai-devsec-gateway --block.
🔒 Deterministic Kill Switch	Hard OS-level blocking through managed hosts entries. No reliance on remote DNS filtering servers.
📦 Universal Distribution	Install via pip, brew, scoop, or as a portable single-file binary for Windows/Linux/macOS.
🌍 Multilingual GUI	A premium Catppuccin Mocha interface with 10 supported languages and smart OS elevation (UAC/sudo).

---

🏢 Enterprise Use Cases

Why do DevSecOps teams and CISOs deploy AI DevSec Gateway?

1. Prevent Data Exfiltration (PII/Secrets): Your team uses Cursor or Copilot, but compliance regulations (GDPR/HIPAA) strictly forbid sensitive environment variables (.env) or proprietary algorithms from leaving the local network.
2. Air-Gapped LLM Routing: You want to transparently force all Copilot traffic within the corporate network to route to an internal, self-hosted Llama-3 (Ollama) server, without requiring developers to change their IDE settings.
3. Shadow AI Auditing: Passively discover which unapproved AI tools or autonomous agents are running in the background on developer workstations.

Read our full Enterprise Use Cases guide for deployment patterns.

---

🎯 Supported Providers

The default interception engine targets 38+ domains across major providers:

Provider	Key domains intercepted
🟢 OpenAI	api.openai.com, chatgpt.com, platform.openai.com
🟠 Anthropic	claude.ai, api.anthropic.com, anthropic.com
🐙 GitHub Copilot	copilot.github.com, api.githubcopilot.com, telemetry.githubcopilot.com
🔵 Google AI	gemini.google.com, aistudio.google.com
🟦 Microsoft	copilot.microsoft.com, bing.com
🔷 Meta AI	meta.ai, ai.meta.com
🌊 Mistral / DeepSeek / xAI	mistral.ai, api.deepseek.com, api.x.ai

The blocklist is dynamically configurable via ai_blocker/constants.py.

---

🏗️ Architecture

AI DevSec Gateway operates at the boundary between your local development environment and the cloud.

graph TD
    subgraph Local Environment ["Local Development Network"]
        IDE[VS Code / Cursor / IDE]
        CLI[ai-blocker CLI]
        GUI[DevSec Gateway GUI]
        
        Gateway((Local API Gateway <br> 127.0.0.1))
        OS_Hosts[(OS Hosts File)]
    end

    subgraph External ["Target Inference"]
        LocalLLM[Private Local LLM <br> Ollama / vLLM]
        PublicCloud[Public Cloud <br> OpenAI / Anthropic]
    end

    IDE -->|1. Outbound API Request| OS_Hosts
    OS_Hosts -->|2. Intercepted| Gateway
    
    Gateway -->|3a. Route active| LocalLLM
    Gateway -.->|3b. Audited pass-through| PublicCloud
    Gateway -->|3c. Block active| Drop[Connection Refused]

For an in-depth dive into our modular structure, Deep Packet Inspection (DPI) plans, and Threat Models, read our Architecture Documentation.

---

✅ Current Capabilities vs Roadmap

This project is intentionally explicit about what is implemented today and what remains future work.

Area	Current status
Hosts-file blocking	Implemented and used by default in GUI/CLI.
Local API gateway	Implemented for loopback HTTP routing to compatible local LLM servers.
Backend selection	Implemented in CLI with hosts default and experimental firewall-redirect dry-run support.
TLS/DPI interception	Implemented. On-the-fly Root CA generator with OS trust store integration for surgical HTTPS inspection.
Token usage monitoring	Implemented. Live /stats endpoint + UI dashboard with auto-refresh.
DLP & guardrails toggle	Implemented. Per-session enable/disable flags for DPI and semantic guardrails.
Expanded DLP patterns	Implemented. Detection of internal IPs (RFC 1918), cloud tokens (GCP, HF, Azure, npm, Slack), database connection strings, and environment variable references.
Enhanced guardrails	Implemented. Leetspeak obfuscation detection and data exfiltration URL patterns (pastebin, gist, webhook).
eBPF/WFP kernel interception	Planned future backend work, not active runtime behavior.

The roadmap is ambitious, but releases should be evaluated by the implemented capabilities above.

---

🚀 Quick Start

1. Python Package (Pip)

The fastest way to get started with the headless CLI.

pip install ai-devsec-gateway

# Native CLI commands are now available globally:
ai-blocker --status
ai-devsec-gateway --block
ai-devsec-gateway --unblock

1.1 Backend Selection & Dry-Run

Use hosts as the default backend, or explicitly inspect the experimental firewall backend with dry-run first:

# Show available backends
ai-blocker --list-backends

# Default behavior (hosts backend)
ai-blocker --backend hosts --block work

# Experimental backend plan only (no network changes applied)
ai-blocker --backend firewall-redirect --block work --dry-run

2. Package Managers (macOS & Windows)

macOS (Homebrew):

brew tap Akunimal/ai-devsec-gateway https://github.com/Akunimal/AI-Router-Blocker-AiO
brew install ai-devsec-gateway
sudo ai-blocker --status

Windows (Scoop):

scoop bucket add ai-devsec-gateway https://github.com/Akunimal/AI-Router-Blocker-AiO.git
scoop install ai-devsec-gateway
ai-blocker --status

3. Portable GUI Binaries

If you prefer a rich visual interface without installing Python:

1. Visit the Releases page.
2. Download the executable for your OS (.exe, macOS binary, or Linux AppImage).
3. Run the application (it will automatically request Admin/sudo privileges when toggling the network switch).

---

🔒 Security Model

• Zero-Persistence BYOK: API keys for the semantic DevSec auditor are strictly kept in-memory. They are never written to disk, preventing supply-chain credential theft.
• Surgical OS Modifications: The engine uses targeted sed-like parsing to inject # AI-Block markers into the OS hosts file. It guarantees absolute isolation from your existing DNS mappings.
• Isolated Telemetry: The application itself contains absolutely zero tracking, analytics, or hidden background phone-home mechanics.

---

🤝 Open Source & Governance

We believe that security tools must be 100% transparent. This project is built under strict open-source governance:

• Architecture Guide: Complete technical specifications.
• Contributing Guide: Standards and PR templates.
• Code of Conduct: We foster a welcoming community.
• Security Policy: Responsible vulnerability disclosure.
• License: MIT Licensed.

---

🗺️ Roadmap & Future Vision

We are actively evolving towards an enterprise Zero-Trust DLP Engine. Upcoming milestones include:

• Cloud-Assisted Semantic DLP: Optional integration with OpenAI API for deep semantic analysis of prompts.
• On-Device Semantic Guardrails: Embed a lightweight ONNX runtime for real-time prompt safety classification.
• eBPF Kernel Telemetry: Detecting .git/config exfiltration directly at the Linux kernel level.
• Confidential Computing: Running the Gateway within Trusted Execution Environments (TEEs) like Intel SGX.

Explore our ROADMAP.md to see the full vision.

---

Audit the unseen. Route the restricted. Trust no packets.
The DevSecOps Gateway for the AI era.