Built Remoroo — an agent for long-running autoresearch workflows

Spaces
1

Hi all,

I built Remoroo, an agent for autoresearch-style workflows and long-running experiments.

The idea is simple: you give it a codebase and a measurable target, and it keeps iterating through the loop instead of stopping at code suggestions:

patch → run → eval → repeat

It’s designed for agentic workflows where:

  • runs take a long time
  • results need to be measured
  • changes should be kept or reverted based on actual outcomes
  • you want a clear record of what was tried

You can try it with:

pip install remoroo
# or
uv tool install remoroo

Then inside a toy repo:

remoroo run --local --engine venv

and follow the prompts.

Links:

If you’re interested in a paid version, sign up first, then email adham@remoroo.com with the subject line Remoroo autoresearch discount and I’ll share a 10% off coupon.

Would love feedback from people here, especially on where this could be most useful in real HF-style workflows.