# How to usefully run 1,000 agents in parallel

1 min read

Last updated 27 Apr 2026

Recorded at Imbue HQ on April 2nd, 2026, [Josh Albrecht](https://x.com/joshalbrecht) (CTO, Imbue) and [Harrison Chase](https://x.com/hwchase17) (CEO, LangChain) discuss how to run a fleet of AI coding agents reliably, safely, and cheaply.

Plus, see a demo of Imbue’s newest tool: `mngr`. A CLI for creating and managing agents locally or in the cloud.

[How to usefully run 1,000 agents in parallel](https://www.youtube.com/watch?v=0BmBs6MPCbw)

With coding agents, what once took months and twenty engineers to build can now be built in a weekend by one. But getting multiple AI coding agents running in parallel—and keeping them running well—is still trickier than it should be.

You’ll come away with:

- A concrete mental model for running parallel agents on real projects
- Insights from two founders building foundational agent infrastructure
- An understanding of `mngr` and how to run 1,000 agents at once

## Get mngr

It’s free, open source, and available today:

```
shcurl -fsSL https://raw.githubusercontent.com/imbue-ai/mngr/main/scripts/install.sh | bash
```

Or get the [code on GitHub](https://github.com/imbue-ai/mngr).