Aideml

AIDE ML is the open source reference implementation of AIDE, a machine learning engineering agent from Weco AI. Given a dataset, a goal and an evaluation metric described in plain English, it drafts, debugs and benchmarks Python code on its own, treating each script as a node in a solution tree and using metric feedback to decide which branches to extend, until the metric is maximized or minimized. It is for agent researchers who want to test new search heuristics, evaluators or model backends, and for machine learning practitioners who want a strong pipeline for a dataset quickly. Each run leaves the best solution script and an HTML visualization of the full search tree, and a Streamlit web UI lets users paste an API key, upload data, set a goal and metric and watch live logs. It works with OpenAI, Anthropic, Gemini or any local model that speaks the OpenAI API. The AIDE algorithm is documented in a published paper and has been used in public research, including OpenAI's MLE bench evaluation. AIDE ML is free under the MIT license and installs with pip; users pay only for their own model usage. The listed website belongs to Weco, whose commercial platform wraps AIDE in a CLI with orchestration, steering mid run, experiment tracking and a dashboard for optimizing models, prompts, agent harnesses and other code against any measurable metric. Weco offers free credits to start and sells additional credits through its CLI, but the credit prices are not published on its public pages. Weco now positions AIDE ML as the research project the company grew out of, and points production users to the Weco platform instead.