<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/i-don-t-want-to-babysit-my-ai-agents-jenazupq2" -->

---
title: I don’t want to babysit my AI agents | daily.dev
description: A founder describes brnrd, an open-source tool built by a two-person team that treats an AI 'resident' as a persistent coworker separate from the underlying...
canonical: https://daily.dev/posts/i-don-t-want-to-babysit-my-ai-agents-jenazupq2
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: I don’t want to babysit my AI agents | daily.dev
og:description: A founder describes brnrd, an open-source tool built by a two-person team that treats an AI 'resident' as a persistent coworker separate from the underlying...
og:url: https://daily.dev/posts/i-don-t-want-to-babysit-my-ai-agents-jenazupq2
og:image: https://api.daily.dev/og/posts/JeNAZuPQ2.png
og:image:alt: I don’t want to babysit my AI agents
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# I don’t want to babysit my AI agents

**[Alexandra Lapunova](https://daily.dev/sources/0f31q6bm6i7mjn7b7o744)** · [@alexandralapunova](https://daily.dev/alexandralapunova) · 5 min read · 2 upvotes · 2 comments

## Summary

A founder describes brnrd, an open-source tool built by a two-person team that treats an AI 'resident' as a persistent coworker separate from the underlying coding agent (Claude Code, Codex, etc). The resident keeps identity, working memory, project context and open work across sessions, so when one agent hits a quota or a process dies, work can continue through another model without losing context. The team has dogfooded it to produce over 1,000 merged PRs in the project's own repository and is inviting community feedback and contributions since the project is still alpha and open source on GitHub.

## Content

brnrd actually started as an idea a couple of years ago.

My co-founder Arseni is a software engineer. I’m very much not one – I’m the CEO and product half of our little operation.

Somewhere between those two perspectives, we kept coming back to the same idea: **what if an AI coworker could actually stay with a project instead of disappearing every time a session ended?**

At the time, we couldn’t really build the thing we had in mind.

Now we can.

Arseni uses coding agents constantly while building our products. Claude Code, Codex, different models for different kinds of work.

And after a while we noticed how much invisible management came with them.

Which agent is better for this task? Who has the right context? What’s still running? What needs restarting? What happens when one hits its quota? How much of the context needs to be carried into the next session?

None of those problems is enormous on its own.

But life is already full of things to manage. We didn’t particularly want managing our AI agents to become another one of them.

So we started thinking about it differently.

Maybe Claude Code or Codex doesn’t have to _be_ the coworker.

Maybe they’re tools the coworker uses.

The thing that persists is the **resident**: its identity, working memory, project context and open work. Claude, Codex, or whatever comes next can be the execution engine underneath it.

That idea eventually became **brnrd**.

**I wanted it to work when I wasn’t “at work”**

One of my favourite use cases is honestly very mundane.

I’m walking the dog. Something about the product pops into my head. Normally I’d either write it down somewhere and hopefully remember it later, or send Arseni another message to add to his infinite pile of things 🤪

Now I can take out my phone, message brnrd and give it the task.

Then I put my phone away and continue walking.

The resident is back on our machine, with the repository and the context it needs, figuring out how to approach it.

And sometimes one of the most interesting parts is coming back and seeing **how** it did something 👀

What did it discover in the project? What tools did it decide to use? What assumptions did it make? What did it change? Did it produce the thing we expected, or arrive somewhere completely different?

There’s something quite fascinating about giving it a goal and later inspecting the trail it left behind.

And there are receipts: commits, branches, PRs and the actual state of the repository. Not just a chat saying, “Done!”

**The process can die. The resident shouldn’t.**

This became particularly obvious once we started moving real work between Claude Code and Codex.

A process can stop. Context can become too large. A provider can hit its quota. A better model for the next part of the task can become available.

Those are properties of the execution layer.

Why should the “person” working on the project disappear with it?

In one of our own runs, Codex hit its quota wall in the middle of ongoing work. The Codex process was done.

The resident wasn’t.

It could continue through Claude with the same working memory and open work.

That distinction – **the resident persists, the harness is replaceable** – has probably become the part of brnrd I find most interesting.

**We made it open**

We also decided fairly early that brnrd should be open source.

Partly because we’re not particularly fond of software asking for meaningful access to our machines and repositories and then essentially saying: _trust us 😇_

If something is going to live on your machine and work with your projects, you should be able to see what it is doing and how it works.

But there’s another reason I care about more.

You should be able to change it.

If our idea of how a resident should behave doesn’t fit your workflow, you should be able to make yours behave differently. If we make a stupid decision, somebody should be able to point at the exact code, disagree with us, and propose something better.

I like the idea of a tool being shaped by the people who actually use it.

Not everything needs to be designed entirely behind closed doors and delivered to a community as a finished answer.

Sometimes the community can participate in deciding what the answer becomes.

**This is still an experiment**

We’ve been dogfooding brnrd rather obsessively.

The resident has already produced **1,000+ merged PRs in brnrd’s own repository**. In a sense, the git history is our longest-running demo.

But it’s still alpha.

There are sharp edges. There are ideas we’ll probably discover were wrong. And I’m sure people will use it in ways we haven’t anticipated at all.

That’s actually part of why I’m writing this here.

I’m curious how people who use multiple coding agents are dealing with this today.

**Do you actively manage every agent?**

Do you pass Markdown files between Claude and Codex? Keep several sessions alive in tmux? Copy context manually? Use an orchestrator? Build your own scripts?

Or do you just accept that when a session ends, the next one wakes up as somebody new?

I’d genuinely like to know.

And if you’re curious about our experiment, brnrd is open here:

[**https://github.com/hugimuni-labs/brnrd**](https://github.com/hugimuni-labs/brnrd)

Look inside. Try it on a real repo. Break it. Tell us which assumptions are stupid. Or change them. 🖤

## Community discussion

Top comments from developers on daily.dev.

**@gurio** · 0 upvotes

> And it is gonna be a mind-blowing game-like agentic dev environment running on your hardware someday 🤓😉

## Similar posts on daily.dev

- [Developers don’t get paid to write code](https://daily.dev/posts/developers-don-t-get-paid-to-write-code-iwue76sb3) · All Things Open · 0 upvotes · 0 comments

---

Tags: [#open-source](https://daily.dev/tags/open-source), [#ai-agents](https://daily.dev/tags/ai-agents), [#claude-code](https://daily.dev/tags/claude-code), [#openai-codex](https://daily.dev/tags/openai-codex)

[View this post on daily.dev](https://daily.dev/posts/i-don-t-want-to-babysit-my-ai-agents-jenazupq2)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/i-don-t-want-to-babysit-my-ai-agents-jenazupq2","headline":"I don’t want to babysit my AI agents","text":"A founder describes brnrd, an open-source tool built by a two-person team that treats an AI 'resident' as a persistent coworker separate from the underlying coding agent (Claude Code, Codex, etc). The resident keeps identity, working memory, project context and open work across sessions, so when one agent hits a quota or a process dies, work can continue through another model without losing context. The team has dogfooded it to produce over 1,000 merged PRs in the project's own repository and is inviting community feedback and contributions since the project is still alpha and open source on GitHub.","url":"https://daily.dev/posts/i-don-t-want-to-babysit-my-ai-agents-jenazupq2","datePublished":"2026-09-17T14:25:58.420Z","dateModified":"2026-09-17T14:26:16.243Z","author":{"@type":"Person","name":"Alexandra Lapunova","url":"https://daily.dev/alexandralapunova","image":"https://media.daily.dev/image/upload/s--4HRzRmOE--/f_auto/v1789679122/avatars/avatar_0f31q6bM6i7MjN7B7o744?_a=BAMAMicg0","description":"🦄 Co-founder \u0026 CEO at HugiMuni · Building brnrd, a persistent resident for coding agents","interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"EndorseAction"},"userInteractionCount":20}},"image":"https://media.daily.dev/image/upload/s--LQpnmKaK--/f_auto/v1789655159/posts/JeNAZuPQ2?_a=BAMAMicg0","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":2}],"comment":[{"@type":"Comment","text":"And it is gonna be a mind-blowing game-like agentic dev environment running on your hardware someday 🤓😉","datePublished":"2026-09-17T14:38:45.448Z","url":"https://daily.dev/posts/JeNAZuPQ2#c-wjcnet9f5","author":{"@type":"Person","name":"Arseni Lapunov","url":"https://daily.dev/gurio","image":"https://lh3.googleusercontent.com/a/ACg8ocI3aN5mjhkomdIIRwsUyTsUh2abkZ1x0GSJseYV8n2cjmcCYkwQmQ=s96-c"}}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/0f31q6bm6i7mjn7b7o744","name":"Alexandra Lapunova"}}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Alexandra Lapunova","item":"https://daily.dev/sources/0f31q6bm6i7mjn7b7o744"},{"@type":"ListItem","position":3,"name":"I don’t want to babysit my AI agents"}]}
```

