Plano (formerly Arch Gateway)

Plano is a self-hosted AI gateway for developers building applications with multiple agents. It handles shared routing, orchestration and monitoring outside the agents, so developers don't need to build those functions into each one.

• Agent orchestration: Coordinate agents written in any language or framework, provided they expose an OpenAI-compatible chat completions endpoint, and add agents without changing application code.
• LLM routing: Send requests to a model by name, use an alias or let Plano select a model based on preferences; LLM routing can run independently of agent orchestration.
• Request visibility: Capture agent activity signals and OpenTelemetry traces and metrics without adding instrumentation code to each agent.
• Moderation and memory hooks: Apply shared filter chains to add jailbreak protection, moderation policies and memory across agents.

Plano is built on Envoy and is open source under Apache 2.0. Its gateway and the Plano family of models, including Plano-Orchestrator, can run locally. A hosted service also runs Plano and its models in the US-central region. Requests sent to that service leave your local deployment.