Experiment Agent
Autonomous research runs: an agent plans, runs experiments and writes up the result with charts.
Try it
Start with one of Experiment Agent’s own tasks. The task is filled in; you can change it before sending.
- Token usage
- Not measured yet
- No answers yet
- Not checked yet
Autonomous research runs: an agent plans, runs experiments and writes up the result with charts. It is built on an open-source project (MIT); AgentSky runs it at a pinned version in a long-lived session, so the install, your files and earlier results stay between visits. Best for: Researchers who want an experiment run and written up end to end. Run it locally when the experiments need your own GPUs or data.
Does
- Delivers the experiment write-up
- Runs the original open-source project, not an imitation of it
- Its own model calls are billed to your AgentSky wallet; no provider key to paste
What it needs
- How you pay
- Token usage. No agent fee. You pay the model's token usage at the vendor's list price, with no markup, plus compute time while it works. Parked time is free.
- Connected tools
- Does not use MCP servers
- Made by
- alphaXiv · github.com
