Plano — AI-Native Proxy and Data Plane for Agentic Apps

Collected: 2026-07-13 20:04 UTC | Topic: AI LLMOps / AI infra

Plano (by Katanemo) is an AI-native proxy server and data plane for agentic applications, built on top of Envoy proxy by its core contributors. It centralizes the “hidden middleware” of agent delivery:

• Agent orchestration: low-latency routing between agents via a purpose-built 4B-parameter orchestrator LLM
• Model agility: route by model name, alias, or automatic preferences through a unified LLM gateway
• Agentic Signals: zero-code capture of traces + OTEL metrics across every agent interaction
• Filter Chains: jailbreak protection, moderation policies, and memory hooks as composable middleware
• Language/framework agnostic: agents are just HTTP servers implementing the OpenAI-compatible chat completions endpoint

Plano operates as an out-of-process sidecar — agents declare themselves in YAML, and Plano handles intent classification, routing, model fallback, provider adaptation, and observability without any code changes to the agents.

Stars: 6,773 | License: Apache 2.0 | Last update: July 2026 GitHub: https://github.com/katanemo/plano