Skip to content
← Tools / Skills
ToolKatanemo/10 trustMetrics-checked

Arch Gateway

Envoy-based intelligent prompt gateway with guardrails for agentic applications.

Visit official site →

About

Arch is an intelligent Layer 7 gateway for agentic applications, built on and by core contributors of Envoy Proxy. It treats prompts like HTTP traffic that needs secure handling, intelligent routing, robust observability, and integration with backend APIs, all outside application business logic.

Purpose-built small LLMs handle function calling and parameter extraction for fast agents and RAG apps, while the Arch Guard model family centralizes jailbreak detection without writing guardrail code. Features include preference-based request routing between upstream LLMs with disaster recovery, prompt targets that wire prompts directly to API endpoints, and unified access to OpenAI, Anthropic, Ollama, and other providers. Caveat: the canonical open-source repo has been quiet since 2025, so treat it as a design reference and check for maintained forks before depending on it.

What it can do

  • Envoy-based Layer 7 prompt gateway
  • Jailbreak and prompt injection guardrails
  • Purpose-built small LLMs for function calling
  • Preference-based routing across LLM providers
  • Prompt targets binding prompts to APIs
  • YAML-driven configuration with observability

Setup

pip install archgw and define an arch_config.yaml with listeners, providers, and prompt targets.

Prerequisites

Python environment; API keys for the upstream LLM providers you route to.

Use cases

agent-gatewaysprompt-guardrailsllm-routingagent-observability

Works with

Trust score — 6/10

/10 trustMetrics-checked

Scored 6/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 91; Trending score 60/100 from researcher assessment.

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.