REA in Locally Uncensored: Let a Local Model Take Apart Apps and Files
Locally Uncensored 3.0.7 has a new plugin for Agent and Code: REA, short for Reverse Engineer Anything. It is an open source MCP server under the MIT license, and it gives your model tools to look inside apps and files, so you can ask "what does this program send, and where?" and get an answer built from the program itself. This guide covers how to switch it on, what it reads, which model to use and what has been tried so far.
REA runs in the desktop app, not in a browser. A full analysis with an answer has run on one Mac so far, with LU's tool code outside the app. On Windows, the built app has started REA and handed its tools to a model, but no full analysis has run there yet. Use REA on software you own or are allowed to examine.
What You Need
- Locally Uncensored 3.0.7, the desktop app
- Node.js 22.19 or newer within 22, 24.11 or newer within 24, or 26 and later. If none fits, the plugin row tells you and names the version it found
- A model of about 27B with a context window of 64K or more. Smaller ones often stop before they write an answer
Switch It On
Open the Plugins menu in the composer of Agent or Code. REA has its own row there and is off by default. When you switch it on, LU fetches the official rea-agents package in version 6.5.0, checks it against the checksum the npm registry publishes, and checks it again before every start. LU starts that exact file and never runs REA's own setup, which would write into the settings of other apps.
The first start downloads REA's dependencies and can take a minute or more. In the built app on Windows, a first start with an empty cache took 62 seconds and brought up 207 tools; a second start, from the cache, took 22 seconds.
207 Tools, but Not All at Once
REA brings 207 tools. Handing all of them to a local model would fill its context before the first question. So a turn only gets the REA tools its words point at. Ask about an Electron app and the model sees the tools for JavaScript and Electron, not the ones for .NET.
Every REA tool sits behind the same confirmation as any external tool in Agent and Code. The model proposes a call, you see it, and it runs once you allow it.
What REA Reads
By itself, REA reads:
- JavaScript and Electron apps
- .NET, Go and Rust programs
- Flutter apps
- SQLite databases
- WebAssembly
Ghidra, Hopper, JADX and Frida are used where you have them installed. LU installs none of them.
Why the Model Size Matters
REA answers with evidence records, and they are big. On a test app of four files, one analysis came back with 343,583 characters, most of it hashes and provenance. A local model cannot read that whole, so LU hands it the facts from each record, about a tenth of the size, together with the id of the record. When a later tool needs that record, the model passes the id and REA looks it up.
Even with that, the work takes several steps and a lot of context. In our run on a Mac, Qwen 3.8 27B on Ollama with 64K context took a small test app apart and wrote a complete answer in 3 minutes 15 seconds. Qwen 3.5 9B with 32K ran out of context before it got there.
Limits
- REA has run on a Mac with LU's tool code outside the app, with Qwen 3.8 27B and Qwen 3.5 9B on Ollama. Only the 27B model wrote a complete answer
- On Windows the built app has started REA and passed its tools to a model; a full analysis there, any run on Linux, and cloud models have not been tried
- REA needs Node.js on the computer; LU does not install it
- Use it on software you own or are allowed to examine
Questions
What is REA in Locally Uncensored?
REA, Reverse Engineer Anything, is an MIT-licensed MCP server. Since 3.0.7 it is a plugin in the desktop app: switched on under Plugins, its tools reach Agent and Code.
What do I need for REA?
The desktop app and Node.js 22.19 or newer within 22, 24.11 or newer within 24, or 26 and later. The plugin row says so when no fitting Node.js is found.
Which model works with REA?
A model of about 27B with a context window of 64K or more. On a Mac, Qwen 3.8 27B on Ollama wrote a complete answer; Qwen 3.5 9B at 32K ran out of context before one.
Does LU install Ghidra or Frida?
No. REA reads JavaScript and Electron apps, .NET, Go and Rust programs, Flutter, SQLite and WebAssembly by itself, and uses Ghidra, Hopper, JADX and Frida only where you already have them.
Can REA run tools without asking?
No. Its tools sit behind the same confirmation as every external tool in Agent and Code.