# How to manage merge conflicts, when running many (50+) agents in parallel — retained record

> Retained reddit record from reddit:ClaudeAI.

Canonical: https://highsignal.app/data/records/19718be3f086ecec

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeAI)  published 2026-11-14 17:52 ingested 2026-11-15 02:34

[original source ↗](https://www.reddit.com/r/ClaudeAI/comments/1xw8nhl/how_to_manage_merge_conflicts_when_running_many/)

## retained text



```
Hey everyone,

I've posted many times about running a ton of agents in parallel, and a popular question I get is "how do I deal with merge conflicts?"

I usually hand waive that away as claude deals with them just fine, just have it rebase etc.

But I've found that an orchestrator agent is the best way to handle many conflicts in parallel (for me at least). This approach has increased my all time high git contributions in a single day from 188, to 320.

https://preview.redd.it/jl705mqb9m1i1.png?width=1820&format=png&auto=webp&s=0505c9336addedbaf0b21828cd79554ebcee85c7

In the screenshot I'm showing 2 orchestrators, but the one I'm focusing on is Lloyd, who is managing 16 separate Opus 5.5 sessions.

When these features start to land, the process of merging will be painful if I decide to do it myself (which I have often done). Not because I have to deal with the conflicts, but because I need to deal with the time spent for each feature to rebase and fix itself before I can review the PR, QA and merge.

Recently though, I decided to take a new approach that mimics how larger dev teams operate. 

Now I ask my orchestrator (Lloyd) to manage a "develop" branch which has all the unreleased / unreviewed (by me) / changes. I tell Lloyd they can merge whatever branches they want into the *develop* branch - as long as its tracked in a plan to help me test and QA the changes when I decide the develop branch is ready to merge to main (which I do manually).

The benefit to this, is it lets the orchestrator work much faster, by moving me as a road block to later down the stream.

The other thing I wanted to illustrate in this screen shot - the orchestrator is communicating with all of the child sessions. You can see them sending messages back and forth. Some people will say this is just a waste of tokens, that I'm spending all my tokens having my agents chat.

But the reality is, this is what's required to scale (for real people and teams, and it maps to agents). 

My agents who are working on features which overlap, all coordinate (through the orchestrator) to make sure they are not creating problems for each other (or to minimize them at merge). The best part of this, is the children agents don't need to message each other - only the orchestrator (which is really a super-powered PM).

I also can treat the orchestrator more like a PM for status updates across the fleet, what's been merged, where are we for the day etc.

How are y'all managing conflicts across many agents?
```



document fetched 2026-11-15 00:21 · canonical https://reddit.com/r/ClaudeAI/comments/1xw8nhl/how_to_manage_merge_conflicts_when_running_many
