Proxiq - Every AI call. Audited. Controlled. Compliant.

The self-hosted LLM gateway security teams demand and developers never notice.

Your Teams → Proxiq (:3099) → Anthropic / OpenAI / Azure / Groq / Ollama / ...

Self-hosted. Open source. No data leaves your infrastructure.

Why teams deploy Proxiq

For your security team:

Block or redact PII before it reaches any LLM — credit cards, SSNs, passports, API keys
Detect and block prompt injection attacks with a configurable threshold
Lock system prompts so no client can override your compliance instructions
Per-user, per-team security policies — Banking, Healthcare, Marketing, or custom
Full audit trail of every request, DLP event, and policy action
SSO via Microsoft Azure AD, Google Workspace, GitHub, Okta, or any OIDC/SAML IdP

For your developers:

Point existing code at localhost:3099 instead of api.anthropic.com — nothing else changes
Drop-in SDK wrapper: relay(new Anthropic()) — one line
65–85% cost reduction via semantic caching, exact caching, and prompt cache injection
Intelligent routing: cheapest capable model tier per request automatically

Install

macOS / Linux — clone and run the setup script (checks all prerequisites, installs Bun if needed, builds and configures):

git clone https://github.com/appmattic/proxiq
cd proxiq
bash setup.sh
./proxiq start

Docker (all platforms including Windows):

docker run -d -p 127.0.0.1:3099:3099 -v proxiq-data:/data ghcr.io/appmattic/proxiq

Then open the dashboard:

http://127.0.0.1:3099/proxiq/dashboard

Your first security policy

Create a named policy and assign it to any user or token. No app changes required.

{
  "policies": {
    "banking-strict": {
      "dlp": {
        "enabled": true,
        "detect": ["credit_card", "ssn", "iban", "api_key", "email"],
        "action": "block"
      },
      "promptGuard": { "enabled": true, "blockThreshold": 0.5 },
      "systemPromptLock": {
        "prepend": "You are a compliant banking assistant."
      },
      "allowedProviders": ["anthropic"],
      "logging": { "storeContent": true, "retentionDays": 365 }
    }
  }
}

Or build policies visually in the dashboard — no config file editing needed.

Cost optimization

Feature	How	Typical saving
Exact cache	SHA-256 match → SQLite hit	100% on repeated queries
Semantic cache	Local MiniLM embeddings, ≥0.94 cosine similarity	40–70% on similar queries
Prompt cache injection	Auto-injects Anthropic `cache_control` breakpoints	~78% on repeated prefixes
Intent routing	Routes each prompt to cheapest capable model tier	30–60% on mixed workloads

SDK integration (one line)

import { relay } from '@proxiq/sdk';
import Anthropic from '@anthropic-ai/sdk';

// All calls route through Proxiq — policies, DLP, caching applied automatically
const client = relay(new Anthropic());

Browse the docs

Getting Started — install, first request, first security policy
Enterprise & Security — DLP, prompt guard, SSO, compliance
Configuration — full .proxiq.json reference
Supported Providers — Anthropic, OpenAI, Azure, Groq, Ollama, and more
SDK Integration — drop-in wrapper for Anthropic/OpenAI clients
CLI Reference — all commands
Claude Connector — MCP setup for Claude Code and Claude Desktop
Deployment — Docker, systemd, AWS, Azure
Architecture — security pipeline and storage design

Enterprise & commercial licensing

Proxiq is open-source under AGPL-3.0. Self-hosting for internal use is always free.

APPMATTIC offers commercial licensing, managed deployment, SLA support, and professional onboarding for regulated industries.

Enterprise details Email APPMATTIC