#179179th of 197 in Coding agents and AI editors

aideml

LLM agent that drafts, runs and iterates ML code via tree search

Stars
1.6k
License
MIT
Last commit
Oct 2026
Last release
Nov 2025
Language
Python
Stars, last 29 weeks+397
21 Apr 2026: 1,222 stars28 Apr 2026: 1,238 stars5 May 2026: 1,246 stars12 May 2026: 1,257 stars19 May 2026: 1,272 stars26 May 2026: 1,283 stars2 Jun 2026: 1,291 stars9 Jun 2026: 1,305 stars16 Jun 2026: 1,313 stars23 Jun 2026: 1,320 stars30 Jun 2026: 1,325 stars7 Jul 2026: 1,334 stars14 Jul 2026: 1,360 stars21 Jul 2026: 1,427 stars28 Jul 2026: 1,446 stars4 Aug 2026: 1,459 stars11 Aug 2026: 1,468 stars18 Aug 2026: 1,478 stars25 Aug 2026: 1,489 stars1 Sept 2026: 1,494 stars8 Sept 2026: 1,506 stars15 Sept 2026: 1,523 stars22 Sept 2026: 1,534 stars29 Sept 2026: 1,546 stars6 Oct 2026: 1,554 stars13 Oct 2026: 1,572 stars20 Oct 2026: 1,585 stars27 Oct 2026: 1,595 stars3 Nov 2026: 1,607 stars10 Nov 2026: 1,619 stars

1,222 on 21 Apr 20261,619 on 10 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

AIDE ML is a Python package that takes a dataset, a plain-English goal and a metric, then runs an LLM-driven tree search over generated scripts, executing and scoring each one. It saves the top-scoring script and an HTML solution tree to logs/. It runs from a CLI, a Python API or a Streamlit UI, and works with OpenAI, Anthropic, Gemini, OpenRouter or local OpenAI-compatible endpoints.

Who it is for: ML engineers and agent researchers automating model-building experiments

Strengths

  • Tree search over code, backed by an arXiv paper and used in MLE-Bench
  • Writes best_solution.py and an HTML solution-tree visualiser per run
  • OpenAI, Anthropic, Gemini, OpenRouter, or local models via OPENAI_BASE_URL
  • MIT license; pip install aideml; CLI, Python API and Streamlit UI

Weaknesses

  • Fully local runs lose quality; feedback model defaults to gpt-4.1-mini
  • Streamlit web UI requires cloning the repo, not just pip install
  • Research reference build; broader features are in the paid Weco product
  • Executes LLM-generated code locally; sandboxing is not described

What it needs

  • Docker
  • Needs LLM API key (OpenAI, Anthropic, Gemini or OpenRouter) or local OpenAI-compatible server, Python 3.10+
  • Models: OpenAI, Anthropic, Gemini, OpenRouter, Ollama / OpenAI-compatible

Quick answers

Is aideml open source?

Yes. Its code is on GitHub under the MIT license.

Can you run aideml with Docker?

Yes. Its repository has a Docker setup: a Dockerfile, a Compose file or a published image.

Also in Coding agents and AI editors

See all 197
Also in Coding agents and AI editors
RankProjectScore
1claw-codeRust CLI coding agent harness using Anthropic and OpenAI-compatible APIs194.8k stars+739 (+0.4%) in 30 daysMIT97 out of 100
2t3codeDesktop, web and mobile front end for controlling local coding agent CLIs Live demo ↗ (opens in a new tab)56.5k stars+29,061 (+106%) in 30 daysMIT94 out of 100
3codewhaleTerminal coding agent in Rust that works with 40+ model providers41.2k stars+334 (+0.8%) in 30 daysMIT94 out of 100
#4piMinimal terminal coding agent extensible with skills and packages121.4k stars+7,392 (+6.5%) in 30 daysMIT92 out of 100
#5oh-my-openagentTerminal coding agent with multi-model orchestration and git-backed memory70.8k stars+1,253 (+1.8%) in 30 dayscustom license92 out of 100