Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Ask HN: What is the "Control Plane" for local AI agents?
7 points by denis4inet 1 day ago | hide | past | favorite | 8 comments
<a href="https://ibb.co/v6QLjdBY"><img src="https://i.ibb.co/S4dV3mxr/Agents-Orchestration.png" alt="Agents-Orchestration" border="0"></a>

I’ve been running an increasing number of local coding agents (Claude Code, Codex CLI, OpenCode, etc.) and I’ve hit a wall: orchestration and state visibility.

When you have multiple agents working on different sub-tasks in a single repo, terminal logs become unmanageable. I find myself needing a "Mission Control" — a centralized tracking system that acts as a bridge between my local terminal sessions and a high-level UI.

The Architecture I’m testing: I’m currently experimenting with using GitHub Issues as a temporary backend for agent state:

On Session Start: The agent hits a hook and creates/updates a GitHub Issue.

On Idle/Output: The agent posts its findings/diffs as a comment.

Human-in-the-loop: I can reply to the issue from my phone/web, and the local CLI picks up the comment to continue or pivot.

The Problem: GitHub Issues (and Jira/Trello) weren't built for the high-frequency event loop of an AI agent. The UX feels sluggish, and there’s no native concept of a "streaming session" or "agent heartbeats."

My Questions to HN:

Is there an emerging SaaS or self-hosted solution that acts as a Jira for Agents?

Are people building custom "Agent Dashboards" that integrate with local CLIs, or is everyone just piping everything to stdout?

If you’re managing 5+ agents working on a codebase simultaneously, how do you track their progress and intervene without context-switching between 5 terminal tabs?

I’ve sketched out a flow where GitHub Issues acts as the hub (linking Codex, Claude Code, and OpenClaw), but I’m looking for something more purpose-built.

Has anyone seen a project that addresses the Control Plane problem for local agents?

 help



I’m building this right now, in a little beyond the POC, but would love your feedback. I have the framework down.

reply


It's not for your use case per se, but at Tacnode we're building what we call a "context lake" that's essentially a living, queryable context layer. It's got way more horsepower than you're describing. Built instead for enterprise-scale needs.

So basically agents always have the most up-to-the-millisecond data freshness they need to make good decisions.

reply


I'm building something that I called control plane for agents, but it's only solving half of the problem you're mentioning. Basically it allows me to connect to my agents remotely from anywhere, so I have one running on my server and several running on my home PC. I generally just use something like notion or beads for issue tracking.

Another thing I do is I have a custom Claude skill that runs every night and goes through all my repositories and Claude conversations and then updates my dashboard and tasks in notion with progress.

https://github.com/amurg-ai/amurg

reply


Works fine for a one-man pet project, but the second you hit a real codebase and dozens of commits a day, that nightly job is just gonna choke on the context window or slam into Anthropic's rate limits. You really need to be streaming agent state into a DB on the fly via WebSockets, not piecing it together once every 24 hours

reply


There's no off-the-shelf "Jira for AI" product out there right now tbh. Most sane teams are either hacking together thin SQLite + WebSockets dashboards, or just sucking it up and staring at raw stdout across 5 tmux panes. If you need a self-hosted setup with tracing, Langfuse should do the trick and cover like 90% of your pain points. Honestly, if you actually polish this control plane concept and open-source it, the demand will be massive

reply


the control plane question is the right one to be asking. most people slap agents into tmux and call it done, then wonder why they cant reproduce what happened last tuesday.

the insight nobody talks about: session identity matters more than session management. if you cant address a specific agent conversation later, search it, or resume it mid-thought, youre just herding cats in a terminal. persistnece + indexing is the real primitive here.

reply


If you're going to post an AI generated response, just say it's AI generated.

reply


probably could be solved partly through langfuse + hooks, if question just in session history

reply




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: