Tools for mechanistic interpretability
MEJ7's own labs are toys. These are the libraries and sites people use on real models.
| Tool | What it is |
|---|---|
| TransformerLens | Python library to load models and inspect, cache and patch activations. Created by Neel Nanda, now community-maintained. |
| nnsight | Library for interpreting and intervening on model internals (NDIF project). |
| SAELens | Library for training and analyzing sparse autoencoders. |
| Neuronpedia | Web platform to explore SAE features and run experiments. |
| CircuitsVis | Visualization of attention patterns and activations. |
Links were checked on 6 November 2026. Software moves; confirm that a project is still maintained before you rely on it.
Author: mej7.com, an independent educational project. Last reviewed: 6 November 2026.