OpenResearch, hosted. Nothing to install.
Autonomous research runs: an agent plans, runs experiments and writes up the result with charts. On AgentSky it runs in a long-lived session around the agent you choose: say what you need and the experiment write-up comes back in the conversation.
OpenResearch has to be cloned, installed and configured on a machine before it does any work, and its results live wherever that machine is.
Choose something else when
Run it locally when the experiments need your own GPUs or data.
How it works
Step 1
It runs the real project
The session clones OpenResearch at a pinned commit and runs its own workflow and scripts, not an imitation of its output.
Step 2
No provider key to paste
OpenResearch's own model calls go through AgentSky and are billed to your wallet, so it works without an account at another model provider.
Step 3
It stays set up
The session is long-lived: the install, your files and earlier results stay between visits, so the next request starts where the last one ended.
What AgentSky runs
- OpenResearch installed in a hosted session
- Its model calls, billed through your wallet
- A session that keeps your work between visits
Powered by OpenResearch
OpenResearch is an open-source project (alphaXiv/OpenResearch, MIT). AgentSky hosts it at a pinned commit; its results come from the project itself.
OpenResearch on GitHub