Hey everyone,
I’ve been working on an open-source, self-hosted AI agent automation platform and wanted to get feedback from people experimenting with agent workflows or local AI setups.
The main idea is to run everything locally (no hosted SaaS) while supporting:
<ol>
<li>multi-step agent workflows</li>
<li>scheduling and retries</li>
<li>execution logs / observability</li>
<li>document chat via RAG</li>
<li>Local model backends (e.g. Ollama-style)</li>
</ol>
I’m less interested in “agent demos” and more focused on the infrastructure side — how agents are orchestrated, monitored, and debugged when they run for long periods.
Things I’d love input on:
<ol>
<li>how others are handling agent orchestration at scale</li>
<li>what’s missing in current open-source agent frameworks for local-first setups</li>
<li>patterns you’ve found useful for logging and failure recovery</li>
</ol>
I’ve documented the architecture and design decisions and would really appreciate any critique, suggestions, or alternative approaches.
<a href="https://github.com/vmDeshpande/ai-agent-automation" target="_blank" rel="noopener nofollow">Github</a>
<a href="https://vmdeshpande.github.io/ai-automation-platform-website/" target="_blank" rel="noopener nofollow">Website</a>
Thanks 🙏

The official Open Source daily.dev Squad. 

Open Source

An open-source, self-hosted AI agent automation platform designed for local deployment without SaaS dependencies. The system supports multi-step agent workflows, scheduling with retries, execution logging and observability, document chat via RAG, and local model backends like Ollama. The project focuses on infrastructure aspects including agent orchestration, monitoring, and debugging for long-running processes. The creator is seeking feedback on orchestration at scale, gaps in current open-source frameworks for local-first setups, and patterns for logging and failure recovery.

Building a fully self-hosted AI agent automation system — looking for feedback