Loopy is a filesystem sandbox whose entire state lives in a single string. It comes with a bash‑style shell so you or your agent can navigate, search, and manipulate a tree‑structured knowledge base with composable commands.
The idea is to simulate a file system to build and navigate a knowledge tree with POSIX-style commands that agents are already familiar with.
Why?
When an agent processes information, it needs somewhere to put it - somewhere it can search, reorganize, and grow organically. For any type of knowledge base like agent memories, product taxonomies, etc the challenge is to expose CRUD type interactions without a pile of specialized tools (search, create, delete, etc.) that are added to context.
Recursive Language Models (RLMs) introduced the idea of putting the entire context into a Python variable and let the model recursively interact with it, instead of reasoning over everything in one shot. RLMs: https://alexzhang13.github.io/blog/2025/rlm/.
I really liked it, but enabling a python REPL seemed like a bad tradeoff for generality.
Loopy imposes a known structure (a tree / filesystem), and replaces the python REPL with a bash syntax over a string. Agents are RL'd in filesystem-like setups, and this provides a sandboxed way to give an agent access to a knowledge base without touching the OS filesystem.
Why this approach:
simple - a single string can represent the full data
known structure - stored in a file system format agents already know and love
composition - compose search commands to quickly navigate the data
Loopy is a filesystem sandbox whose entire state lives in a single string. It comes with a bash‑style shell so you or your agent can navigate, search, and manipulate a tree‑structured knowledge base with composable commands.
The idea is to simulate a file system to build and navigate a knowledge tree with POSIX-style commands that agents are already familiar with.
Why? When an agent processes information, it needs somewhere to put it - somewhere it can search, reorganize, and grow organically. For any type of knowledge base like agent memories, product taxonomies, etc the challenge is to expose CRUD type interactions without a pile of specialized tools (search, create, delete, etc.) that are added to context.
Recursive Language Models (RLMs) introduced the idea of putting the entire context into a Python variable and let the model recursively interact with it, instead of reasoning over everything in one shot. RLMs: https://alexzhang13.github.io/blog/2025/rlm/.
I really liked it, but enabling a python REPL seemed like a bad tradeoff for generality.
Loopy imposes a known structure (a tree / filesystem), and replaces the python REPL with a bash syntax over a string. Agents are RL'd in filesystem-like setups, and this provides a sandboxed way to give an agent access to a knowledge base without touching the OS filesystem.
Why this approach:
simple - a single string can represent the full data known structure - stored in a file system format agents already know and love composition - compose search commands to quickly navigate the data