#5252nd of 52 in Agent platformsFor science, health and research, research and papers
ai-scientist-v2
Agentic tree-search pipeline that generates ML experiments and writes research papers
- Stars
- 7.4k
- License
- custom license
- Last commit
- Dec 2025
- Language
- Python
5,275 on 10 Apr 20267,427 on 30 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Takes a research topic in Markdown, brainstorms ideas with literature-based novelty checks, then runs a tree search over LLM-written experiment code. It analyzes results and writes a LaTeX paper PDF, with an LLM review step. Unlike v1 it needs no human-authored templates, but the README says success rates are lower than v1 with a strong template.
Who it is for: ML researchers experimenting with automated, open-ended paper generation
Strengths
- End-to-end pipeline: ideation, experiments, plots, LaTeX write-up and review in one repo
- Tree search settings (workers, steps, debug depth, drafts) are configurable in bfts_config.yaml
- Supports OpenAI, Gemini and Claude via AWS Bedrock; README gives per-run cost estimates
- No human-written templates required, unlike v1
Weaknesses
- Executes LLM-written code; README tells users to sandbox it themselves, no Docker provided
- Custom license (NOASSERTION) requires disclosing AI use in resulting manuscripts
- Linux with NVIDIA GPU and CUDA assumed; installation can take up to an hour
- Lower success rate than v1; runs often finish without a PDF
What it needs
- GPU required
- Needs PyTorch with CUDA, conda, poppler, chktex, LaTeX tools, Semantic Scholar API (optional)
- Models: OpenAI, Gemini, Claude via AWS Bedrock
Quick answers
Is ai-scientist-v2 open source?
Its code is public on GitHub under its own license terms; read them before you use it at work.
Does ai-scientist-v2 need a GPU?
Yes, it needs a GPU.
Also in Agent platforms
See all 52| Rank | Project | Score |
|---|---|---|
| 1 | AutoGPTRole-based AI agents coordinated by one lead agent, plus a visual workflow builder | 94 out of 100 |
| 2 | LangflowVisual builder for AI agents and workflows, exposed as API or MCP server | 93 out of 100 |
| 3 | daily_stock_analysisLLM-driven daily analysis of watchlist stocks across six markets, pushed to chat apps | 93 out of 100 |
| #4 | DifyVisual builder for LLM workflows, RAG pipelines and agents, with APIs | 92 out of 100 |
| #5 | open-designDesign workspace that drives your coding agent to build prototypes, decks and video | 92 out of 100 |