Plano (formerly Arch Gateway)

// Overview

Plano (formerly Arch Gateway) is an out-of-process proxy server and data plane for agentic applications. Built on Envoy by its core contributors, it moves the repetitive plumbing of production agents - routing between agents, guardrail and moderation hooks, LLM access, and observability - out of your framework and into a separate data plane you configure with YAML.

It is aimed at teams who can build an agent demo quickly but struggle to ship it safely and repeatably. Instead of writing intent classifiers, model fallbacks, provider adapters, and tracing glue in every codebase, you declare your agents and model providers once and let Plano handle the wiring. Your agents stay as plain HTTP services in any language or framework.

As an LLM gateway and proxy, Plano sits between your services and the models. It can route by model name, by alias, or automatically by preference, capture OTEL traces and metrics with no extra code, and apply moderation and memory policies through filter chains.

// What it does

• Low-latency orchestration between agents - add new agents without changing application code
• Smart LLM routing: route by model name, semantic alias, or automatically by preference
• Zero-code capture of agentic signals plus OpenTelemetry traces and metrics across every agent
• Guardrail filter chains for jailbreak protection, moderation policies, and memory consistency
• Unified, OpenAI-compatible access to multiple LLM providers (OpenAI, Anthropic, and more)
• Built on Envoy as a separate out-of-process data plane, so it works with any language or framework

// Getting started

Plano runs as a separate proxy configured by a YAML file. You declare your agents and model providers, run your agents as plain HTTP services, then start Plano and query it.

# config.yaml
version: v0.3.0

agents:
  - id: weather_agent
    url: http://localhost:10510
  - id: flight_agent
    url: http://localhost:10520

model_providers:
  - model: openai/gpt-4o
    access_key: $OPENAI_API_KEY
    default: true
  - model: anthropic/claude-3-5-sonnet
    access_key: $ANTHROPIC_API_KEY

listeners:
  - type: agent
    name: travel_assistant
    port: 8001
    router: plano_orchestrator_v1
    agents:
      - id: weather_agent
        description: |
          Gets real-time weather and forecasts for any city worldwide.
      - id: flight_agent
        description: |
          Searches flights between airports with live status and schedules.

tracing:
  random_sampling: 100

# Start Plano
planoai up config.yaml

# Query - Plano routes to the right agent
curl http://localhost:8001/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-4o"}'

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Routing a single conversation across multiple specialized agents (for example a weather agent and a flight agent) without hard-coding routing logic
• Giving services unified, OpenAI-compatible access to several LLM providers with automatic model fallback
• Adding jailbreak protection, moderation, and memory policies to an agentic app through filter chains instead of bespoke code
• Capturing traces, metrics, and agentic signals across all agents for evaluation and continuous improvement

// How Plano (formerly Arch Gateway) compares

Plano (formerly Arch Gateway) alongside other open-source gateways & routing tools AI/TLDR tracks, ranked by GitHub stars.