scan_codeLABSCO SUMMARY
A map of the codebase for the price of one call.
Handing a model a repository usually means it reads files until the context window fills. A definition list with line numbers gives it the structure first, so the files it then opens are chosen rather than stumbled into — and the filters are what make that work on something large, since scanning a monorepo unfiltered just moves the problem. Output format matters more than it sounds: XML and JSON are there because a model parses them more reliably than prose, and Markdown is there because you might want to read it yourself.
What it is
A structural scanner rather than a text search. It parses source files with Tree-sitter and reports the definitions it finds, so an assistant can get the shape of a codebase without reading every file — and the output formats are chosen for a model to consume.
What you get
- One tool that scans a directory and returns its definitions — functions, classes, variables, interfaces and methods — with line numbers attached —
scan_code - Six languages through Tree-sitter: JavaScript, TypeScript, C#, PHP, CSS and Python, matched by extension
- Rules from your
.gitignorerespected automatically, so vendored and build directories do not swamp the result - Filtering that actually narrows a large repository: by definition type, by modifier such as public or private, by a regex on names, and by glob patterns on file paths — each available as both an include and an exclude
- Three output formats — Markdown by default, or XML or JSON — and three verbosity levels from minimal through standard to detailed
- The same functionality as a CLI: pass a directory argument and it runs standalone, omit it and the same binary starts as an MCP server on stdio
Requirements
No account and no key. Node.js and npm: clone, npm install, npm run build, and point your client at the built entry with node using an absolute path. The package is code-scanner-server (0.1.0 in package.json). The directory to scan is a required argument on every call, so the server is not bound to one project. The README notes the tool is under active development and that some language parsers are less tested than others.
Setup effort
Build from source — git clone https://github.com/Ixe1/code-scanner-server.git
USE CASES
01
Getting oriented in an unfamiliar repository
Scan it once for the definition list with line numbers, and use that as the map before opening any file.
02
Finding the public surface of a module
Filter to classes and methods with a public modifier, so the result is the API rather than every internal helper.
03
Tracking down a naming convention
Filter definitions by a regex on the name across the whole tree, and see every place the pattern is used.
04
Keeping the scan small on a big repository
Restrict it to the paths and file patterns you care about and drop the detail level, so the result fits in a reply.
1 tool
Listed in project docs. Your client may show a different list.
BEFORE YOU INSTALL
What to know first
Tools you get
1 tool named in the project's documentation.
Where it runs
Both documented clients run the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
git clone https://github.com/Ixe1/code-scanner-server.git
cd code-scanner-server
npm install
npm run buildMCP Clients
Claude Desktop, Cline
JSON
{
"mcpServers": {
"code-scanner-server": {
"command": "node",
"args": [
"<YOUR_PATH_TO_CODE_SCANNER_SERVER>/build/index.js"
],
"env": {},
"disabled": false,
"autoApprove": []
}
}
}
