AI Frameworks · katanemo

plano

Plano is an AI-native proxy and data plane built in Rust that centralizes agent orchestration, LLM routing, observability, and safety guardrails for agentic applications. It runs as an out-of-process service that agents communicate through, eliminating the need to rewrite common production infrastructure in every codebase.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/katanemo/plano

6.7k

GitHub stars

443

Forks

Rust

Primary language

Apache-2.0

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	katanemo/plano
Owner	katanemo
Primary language	Rust
License	Apache-2.0 — OSI-approved
Stars	6.7k
Forks	443
Open issues	132
Latest release	0.4.26 (2026-06-25)
Last updated	2026-07-07
Source	https://github.com/katanemo/plano

What plano is

Built on Envoy by core Envoy contributors, Plano provides HTTP-based agent routing via semantic intent classification (using lightweight LLMs like a 4B-parameter orchestrator), OpenTelemetry tracing instrumentation without code changes, filter chains for moderation/safety, and unified LLM provider APIs (OpenAI, Anthropic) with model aliasing and failover logic.

Quickstart

Get the plano source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/katanemo/plano.git2cd plano3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Multi-agent orchestration at scale

Route user intents to specialized agents (weather, flight, research) without writing custom classifiers. Plano's 4B routing model handles intent detection; add agents via YAML config without redeploying application code.

Unified LLM provider abstraction

Centralize model routing, failover, and cost optimization across OpenAI, Anthropic, and other providers. Switch models or add fallbacks in config rather than updating application logic.

Production observability for agents

Automatic end-to-end tracing with OpenTelemetry and structured 'Agentic Signals' for evaluation and continuous improvement without instrumenting every service or agent individually.

Implementation considerations

Requires HTTP-compatible agents; agents must implement OpenAI-compatible `/v1/chat/completions` endpoint for Plano to route through them.
YAML configuration for agents, models, and listeners; changes typically require service restart (exact hot-reload behavior not documented).
Free hosted Plano LLMs (routing models) in US-central; production scaling requires running orchestrator LLMs locally or obtaining API keys from vendor.
OpenTelemetry integration is automatic but requires exporter configuration for traces to reach observability backend (e.g., Jaeger, Datadog).
Rust-based binary; deployment via Docker or native binary; no managed SaaS offering documented in README.

When to avoid it — and what to weigh

Lightweight, single-agent applications — Adding an out-of-process proxy adds latency and operational overhead. Simple single-agent systems may be over-engineered by Plano's full feature set.
Strict latency budgets (<50ms per hop) — Plano introduces network hops between agents and the proxy. Real-time, ultra-low-latency scenarios may see unacceptable delay increases.
Proprietary or custom LLM providers only — Plano's model routing and unified APIs target mainstream providers. Highly custom or internal-only LLM infrastructure may require significant extension work.
Existing heavy investment in competing frameworks — If you have mature tooling in LangChain, AutoGen, or similar, migration effort and dual-maintenance complexity may outweigh benefits for small teams.

License & commercial use

Apache License 2.0 (Apache-2.0). Permissive OSI-approved license permitting commercial use, modification, and distribution with attribution and liability disclaimer.

Apache-2.0 is permissive and allows commercial use of the software itself. However, the README states that Plano-hosted orchestrator LLMs are 'free of charge' for US-central region only; production deployment at scale requires either running LLMs locally or contacting the vendor for 'API keys,' implying a commercial arrangement. Commercial terms for orchestrator model access require vendor discussion.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Strong
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Strong
Assessment confidence	High

Security considerations

Filter chains support moderation and jailbreak protection policies, but details on built-in filters, custom extension APIs, and validation rigor are not documented. Network security relies on standard HTTP/TLS practices between Plano and agents; no mention of mutual TLS or agent authentication. Free hosted LLMs in US-central region may be subject to shared-infrastructure risks. OpenTelemetry traces may contain sensitive model inputs/outputs; exporter configuration and retention policies require careful review to avoid leaking PII.

Alternatives to consider

LangChain / LangSmith

Mature Python/JS agent framework with built-in observability (LangSmith) and multi-model support. Better if you prefer framework-embedded orchestration; less modular than Plano's out-of-process design.

Anthropic Workbench / Claude API directly

If standardizing on Anthropic Claude and extended thinking, native tooling avoids proxy latency. Narrower scope than Plano's multi-provider routing.

OpenAI Swarm / AutoGen

Lightweight agent coordination libraries; no out-of-process proxy. Easier to embed in existing applications but less separation of concerns than Plano.

Software development agency

Build on plano with DEV.co software developers

Plano removes routing, orchestration, and observability boilerplate. Try the quickstart, explore the demos, or join the Discord community to see how it accelerates your agent workflows.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

open-multi-agent ai-gateway plandex

Related on DEV.co

Explore the category and the services that help you build with it.

AI Frameworks AI Application Development API Development DevOps & Deployment Talk to DEV.co

plano FAQ

Do agents need to know about Plano, or is it transparent?

Agents implement standard OpenAI-compatible `/v1/chat/completions` endpoints and call back to Plano's LLM gateway (http://localhost:12001/v1) for model inference. Plano is not transparent to agents; they must be aware and configured to use Plano's gateway.

Can I run the routing/orchestrator models locally instead of using hosted versions?

README confirms production deployments can 'run these LLMs locally.' Exact instructions for local model deployment (model weights, hardware requirements) not provided in README; likely in full documentation.

What languages can agents be written in?

Any language that can expose an HTTP `/v1/chat/completions` endpoint. README examples show Python (FastAPI) but framework-agnostic. Go, Node.js, Java, etc., are viable.

Is there a managed/SaaS version of Plano?

Not documented in README. Free hosted models in US-central for development only. Production likely requires self-hosted deployment or vendor-managed API keys (requires contact/signup).

Work with a software development agency

Need help beyond evaluating plano? DEV.co is a software development agency offering software development services and web development for teams of every size. Our software developers and web developers build custom software, web applications, APIs, and ai frameworks integrations — and maintain them long-term.

Ready to simplify agentic application delivery?

Plano removes routing, orchestration, and observability boilerplate. Try the quickstart, explore the demos, or join the Discord community to see how it accelerates your agent workflows.

Talk to DEV.co Schedule a Consultation