Adyant

Smart Markov-chain URL wordlist generator for fuzzing and recon.

Instead of brute-forcing targets with massive, static, and noisy wordlists, Adyant learns from real URL patterns (via Burp history, Wayback, etc.) and generates statistically likely paths tailored specifically to your target.

If you want to find hidden endpoints with fewer requests and less noise, you're in the right place.

Remeber: Garbage in = Garbage out; feed it well. the model shown above was trained on a courpus of randomly collected URLs from bigquery. Use your own.

we're taking log of probabilities, so the result is a negative value, don't worry about it.

---

⚡ Quickstart (Time-to-First-Value)

Get Adyant running locally in under 30 seconds.

1. Install via pip

pip install adyant

Recommended

git clone https://github.com/forshaur/adyant.git
cd adyant
pip install .

2. Train & Fuzz (Seamless Pipeline)

Feed it URLs, give it a seed, and pipe it directly into your favorite fuzzer:

Bash

cat burp_urls.txt | adyant --train - --seed https://target.com/api/ -q --paths-only | ffuf -u https://target.com/FUZZ -w -

✨ The Value Proposition

Why replace your static wordlists with Adyant?

• 🧠 Context-Aware Fuzzing: It doesn't just guess /admin; it calculates the probability of /admin/v2/metrics based on real-world transition states.

• 🎯 Multiple Attack Modes: Need the most obvious paths? Use sample. Looking for deeply nested routes? Use deep. Hunting for hidden gems? Use rare.

• 🛠️ Native Integrations: Outputs natively formatted payloads for ffuf, burp, and nuclei with zero parsing scripts required.

• 🤖 Smart Synonym Discovery: Optional ML-driven clustering (via scikit-learn & sentence_transformers) identifies semantic domain clusters and token synonyms automatically.

• 📉 Reduced Request Volume: Higher hit rates with a fraction of the HTTP requests. Stay under WAF rate limits.

📖 Documentation & Usage

Adyant is highly tunable. You can adjust the Markov context (--context), control rarity (--rarity), and format outputs to include explicit confidence scores (--scores).

For advanced configurations, saving/loading models, and a detailed breakdown of all generation modes, please refer to the official documentation:

👉 Read the full Adyant Wiki

🤝 Community & Support

• Issues & Bugs: Encountered an error? Please open an issue.

• Research: This project is backed by ongoing security research, which will be published and linked here shortly.

⭐️ Don't forget to star it so that you may use it later.