Plano

Maker here 👋 We built Plano, because building agentic demos was easy, but delivering agentic apps safely, reliably, and repeatably to production is very hard. Plano is delivery infrastructure for agentic apps - an AI-native proxy server and data plane designed to help you build agents faster, and deliver them reliably to prod.

The reality of most agentic development is that after a quick hack, you end up building or maintaining the “hidden AI middleware” to reach production: routing logic to reach the right agent, guardrail hooks for safety and moderation, evaluation and observability glue for continuous learning, and model/provider quirks — scattered across frameworks and application code.

Plano solves this by moving core delivery concerns into a unified, out-of-process dataplane. Core capabilities:

• 🚦 Orchestration: Low-latency orchestration between agents, and add new agents without changing app code. When routing lives inside app code, it becomes hard to evolve and easy to duplicate. Moving orchestration into a centrally managed dataplane lets you change strategies without touching your agents, improving performance and reducing maintenance burden while avoiding tight coupling.
  

• 🛡️ Guardrails & Memory Hooks: Apply jailbreak protection, content policies, and context workflows (e.g., rewriting, retrieval, redaction) once via Filter Chains at the dataplane. Instead of re-implementing these in every agentic service, you get centralized governance, reduced code duplication, and consistent behavior across your stack.
  

• 🔗 Model Agility: Route by model, alias (semantic names), or automatically via preferences so agents stay decoupled from specific providers. Swap or add models without refactoring prompts, tool-calling, or streaming handlers throughout your codebase by using Plano’s smart routing and unified API.
  

• 🕵 Agentic Signals™: Zero-code capture of behavior signals, traces, and metrics consistently across every agent. Rather than stitching together logging and metrics per framework, Plano surfaces traces, token usage, and learning signals in one place so you can iterate safely.

Built by core contributors to the widely adopted Envoy Proxy <https://www.envoyproxy.io/>_, Plano gives you a production‑grade foundation for agentic applications. It helps developers stay focused on the core logic of their agents, helps product teams shorten feedback loops for learning, and helps engineering teams standardize policy across agents and LLMs. Plano is grounded in open protocols (de facto: OpenAI‑style v1/responses, de jure: MCP) and proven patterns like sidecar deployments, so it plugs in cleanly while remaining robust, scalable, and flexible.

Would love feedback from folks shipping agents in production: what’s the most painful bit of middleware you’ve had to rebuild (routing, safety, traces, state, tool glue, eval loops)?

Cool Demonstration, Is this more on the safe-and-reliable side of agents, with similarities to superagents by clickup? I'm sure you saw their launch, what's Plano's strenght against that?

The separation of delivery concerns from agent logic makes sense for production deployments. I'm curious about the Filter Chains approach—does Plano support async guardrail processing for latency-sensitive agents, or is it primarily synchronous?