Show and Tell: Backplaned — A multi-agent orchestrator

[SJK-py]

Hi everyone,

I just finished the alpha stage of a new project called Backplaned, and I wanted to share it here because Lemonade is acting as the local LLM engine driving the entire testing system and it is rock solid!

What is Backplaned?
It’s a lightweight, self-hosted multi-agent orchestration platform. Instead of a monolithic app, it uses an ESB-style central router to handle task routing, access control, and proxy files between pluggable agents over a unified HTTP protocol. It ships with built-in agents for sandboxed code execution, document-to-markdown conversion, long-term memory, and MCP bridges.

Where Lemonade Fits In:
Running complex multi-agent loops, parallel tool calling, and autonomous research requires a rock-solid inference backend. I've been running Lemonade to serve my local LLMs on my Strix Halo setup.

Because the Lemonade backend handles multiple local models efficiently, Backplaned’s LLM agent can smoothly distribute inference workloads to the multiple models with different size, capability, and latency.

If you are interested in local LLM powered agent, check out the repo!

Repo link: https://github.com/SJK-py/backplaned