# Plano

An AI-native proxy server and data plane for agentic applications that centralizes LLM routing, agent orchestration, observability, and guardrails.

Plano sits between an application and its language models as an out-of-process data plane, so routing, orchestration, and safety logic do not have to be rebuilt inside every framework or codebase. It routes by model name, alias, or automatic preference, captures agent signals and OpenTelemetry traces and metrics without code changes, and applies moderation policies and memory consistently through filter chains.

It is built on Envoy and is intended to work with any programming language or AI framework, letting teams add new agents without modifying application code. A hosted set of Plano models is available for a first run, with the option to run models locally or supply API keys for production use.

Plano is written in Rust and distributed under the Apache-2.0 license.

Website: https://planoai.dev