Our take
Newless than six months old- Activity
- 32
- Community
- 4
- Issues
- 65
- Pull requests
- 61
-22 stars in the last 30 days
- Fewer contributors than 91% of the projects we track
AutoResearch offers a traceable research workflow, but it is a new project whose development and model configuration currently depend on one contributor.
Good fit if
- You need an end-to-end workflow for AI or machine-learning research and want experiment records and independent reviews saved locally.
- You can configure compatible model endpoints, including three distinct models for idea generation, and provide suitable experiment compute.
- You are comfortable evaluating a new Apache-2.0 project whose recent activity includes commits and merged changes, but whose work is concentrated in one contributor.
Look elsewhere if
- You need a mature project with a broader contributor base or an established release history.
- You need a turnkey research service that does not require API credentials, model routing, and a separate execution environment.
- Your research does not fit an agent workflow for AI and machine-learning experiments.
All health signals
| Last commit | 2026-09-13 (2 months ago) |
|---|---|
| Commits, last 90 days | 7 |
| Releases, last 12 months | 0 |
| Contributors | 1 |
| Issues closed, last 90 days | 3 (typically closed in 1 day) |
| Pull requests merged, last 90 days | 4 (typically merged in 0 days) |
| Project age | 3 months |
Checked on 2026-11-07 with the GitHub API.
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
3,102Forks
279Change since Nov 16, 2026: +2 stars·0 forks
Overview
AutoResearch is EvoMap’s Python-based agent workflow for AI and machine-learning research. It takes a supplied research idea, or directions discovered from recent papers, developer communities, and open-source trends, through experiment planning, implementation, execution, analysis, and independent evaluation. The aim is to turn an idea into traceable, paper-ready evidence rather than an unsupported model-generated conclusion.
It is intended for researchers and technical teams who can configure model services and provide the resources an experiment needs. The workflow saves plans, code, logs, results, and reviews to disk so work can resume and people can inspect or take over; it does not guarantee that conclusions are correct.
Key Features
- Collects and filters external research signals, then combines accepted directions with a user-maintained local knowledge base.
- Uses at least three distinct models for idea generation and cross-review.
- Produces experiment plans that can be selected and exported with provenance records.
- Runs experiments through a stateful workflow with pilot runs, recoverable work queues, and saved project state.
- Records run logs, metrics, failure causes, and decisions, including negative results.
- Adds critic and blind-review stages to challenge results independently.
- Provides project and knowledge-base dashboards for inspecting saved work.
Use Cases
- A research group with no settled topic can use idea generation to combine current research signals with its own domain knowledge.
- A researcher with a hypothesis can start directly from an idea file and use the execution workflow to plan, pilot, and evaluate experiments.
- A team comparing candidate directions can preserve sources, plans, and review records for later inspection.
- Researchers running long experiments can resume from saved project state after interruption and inspect failures as well as successful results.
What you need
Detected in the repository
- Python (from requirements.txt)
- A test suite and automated checks on GitHub Actions
License in plain words
Apache-2.0permissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the license, the NOTICE file and a note of your changes
- Share your changes: no
- Includes an explicit patent grant from the contributors:
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
On Linux or an SSH machine with Git and Python 3.10+, clone the repository and run its setup script:
git clone https://github.com/EvoMap/AutoResearch.git
cd AutoResearch
bash scripts/bringup.sh
The setup script installs dependencies and runs baseline checks. Configure model credentials and routes before running the pipeline. See the README for configuration and execution details.
Considerations
- The repository is new, has no release history in the measured period, and its code contributions are concentrated in one person. Recent commits and merged pull requests show activity, but do not establish long-term maintenance.
- The research workflow relies on external model services and can make real API requests. Idea generation requires at least three distinct underlying models, and usage can incur charges.
- Idea execution has additional requirements: Bun, Node.js, an experiment environment manager, and the CPU, GPU, data, and disk resources required by the experiment. GPU needs depend on the experiment, not the idea-generation stage.
- The README warns that the current Alpha coordinator uses broad tool permissions. It should run in an isolated, disposable environment without access to unrelated credentials or data.
- AutoResearch preserves evidence for review but explicitly does not guarantee that every conclusion is correct.
Found this useful?
Share it with someone who would like AutoResearch.