Our take
Activerecent releases- Activity
- 45
- Community
- 45
- Issues
- 21
- Pull requests
- 43
+5 stars in the last 30 days
subwiz is a focused way to add model-generated candidates to subdomain discovery, with recent commits and releases and a contributor base that is not concentrated in one person.
Good fit if
- You already have known subdomains and want additional candidates to investigate.
- You want a CLI or Python interface and can run a Python 3.9–3.13 application.
- You want to choose between CPU, CUDA, or MPS inference and control whether predictions are resolved.
Look elsewhere if
- You need a tool that discovers subdomains without seed input or replaces passive enumeration.
- You require a documented accuracy benchmark or guaranteed valid discoveries; none is provided in the supplied project information.
- You need a different runtime, as the stated Python support is limited to versions 3.9 through 3.13.
All health signals
| Last commit | 2026-09-24 (21 days ago) |
|---|---|
| Commits, last 90 days | 3 |
| Releases, last 12 months | 2 (latest v1.0.1, 2025-11-11) |
| Contributors | 6 (top contributor: 55% of commits) |
| Issues closed, last 90 days | 0 |
| Pull requests merged, last 90 days | 1 |
| Project age | 2 years |
Checked on 2026-10-16 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
394Forks
27Change since Nov 14, 2026: 0 stars·0 forks
Overview
subwiz is a lightweight subdomain discovery tool built on nanoGPT, the transformer project by Karpathy. It uses a model trained on lists of subdomains from passive sources to predict additional names from subdomains you already know, offering another source of leads alongside tools such as subfinder.
It is aimed at security researchers and developers assessing a domain's exposure. The Python CLI runs on Python 3.9 through 3.13 and supports automatic device selection as well as CPU, CUDA, and Apple MPS; it can also be called from Python code.
Key Features
- Generates candidate subdomains from a newline-separated list of known subdomains.
- Uses beam search to produce multiple likely sequences rather than just one output.
- Can recursively use newly found subdomains as input for further predictions.
- Resolves predictions by default, with an option to skip resolution and inspect raw model output.
- Filters wildcard DNS results or skips apex domains with wildcard records.
- Supports multiple apex domains in one input file with the
--multi-apexoption. - Offers CLI output controls for files, quiet piping, and silent operation, plus a Python API.
Use Cases
- A penetration tester can seed subwiz with passive results from subfinder to generate additional candidates for an authorized assessment.
- A bug bounty researcher can explore likely subdomain names when passive enumeration has produced an initial set of seeds.
- A security team can inspect unfiltered model predictions with
--no-resolvewhen DNS resolution would be undesirable or premature. - A developer can integrate predictions into a Python workflow that already collects known subdomains.
What you need
Detected in the repository
- Python >=3.9, <3.14 (from pyproject.toml)
- A Dockerfile, so it can run in a container
- A test suite and automated checks on GitHub Actions
License in plain words
MITpermissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the copyright and license notice
- Share your changes: no
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
Install with pipx and seed it with results from a passive discovery tool:
pipx install subwiz
subfinder -d example.com -o subdomains.txt
subwiz -i subdomains.txt
See the README for all CLI options and Python usage.
Alternatives
- mantis: Mantis combines asset discovery with vulnerability and configuration scanning, while subwiz predicts subdomains from known ones and checks their DNS responses.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| subwiz | Python | MIT | 389 | Active |
| mantis | Python | Apache-2.0 | 1k | Quiet |
Considerations
The model proposes likely names; predictions should be treated as candidates, not confirmed subdomains. Resolution is enabled by default, and the README describes wildcard DNS handling, but does not provide accuracy or coverage measurements. The project supports CPU as well as CUDA and MPS, so a GPU is not listed as a requirement. It has recent activity and releases, though recent issue-closure data shows no issues closed in the measured period. The MIT license permits broad use subject to its license terms.
Found this useful?
Share it with someone who would like subwiz.