numbers-le-mcp
Extracts numeric literals from C++ source code, including hex, binary, octal, digit separators, and suffixes, with notation and position metadata where available.
Extracts numeric values from environment/.env files when a value is entirely numeric, such as PORT=3000.
Extracts numeric literals from JavaScript source code, preserving notations such as decimal, hex, binary, octal, scientific, and bigint.
Extracts numeric literals from Kotlin source code, respecting Kotlin numeric notations and suffixes.
Extracts numeric literals from Python source code, including hex, binary, octal, digit separators, and suffixes.
Extracts numeric literals from Rust source code, including hex, binary, octal, digit separators, and suffixes.
Extracts numeric literals from shell scripts using the numeric-literal extractor.
Extracts integer and float values from TOML documents; dates are not treated as numbers.
Extracts numeric literals from TypeScript source code, preserving numeric notation and position metadata where available.
Extracts numeric values from YAML documents, including hex/octal integer forms and exponents, across multi-document streams.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Useful? A star or rating is how other developers find it — ★ GitHub · ★ Open VSX · ★ Marketplace
What it does
Open a file, run Numbers-LE: Extract Numbers, and every numeric value in the document lands in a new editor — deduplicate and sort it from there. Works in VS Code and in VS Code–based editors like Cursor and VSCodium (installable from Open VSX).
Data validation — pull the numbers out of a config or fixture and eyeball ranges at a glance
Config audits — compare ports, thresholds, and limits across INI/TOML/.env files
CSV work — extract everything, one column, or several columns into separate documents
Related MCP server: paths-le-mcp
Install
Where | What you get | Install |
VS Code | The same extraction, in your editor, on a keystroke | |
Cursor, VSCodium, Windsurf | The same extension | |
A terminal or a CI step | The same run over a whole tree, with exit codes |
|
Any MCP agent, via Node |
|
|
Use it from an AI agent
The same engine runs as an MCP server, so an agent can call it directly instead of you running a command.
Editor | How |
VS Code 1.101+ | Nothing to install — the extension registers |
Claude Code |
|
Cursor, Windsurf, anything else | point it at |
extract_numbers(content, format?, filename?, dedupe?, maxResults?)Returns the values in document order, each with the notation it was written in and — for JSON, source code and plain text, where the position is exact — its 1-based line and column; a parsed format (YAML, TOML, CSV, INI, dotenv) carries no position rather than a guessed one. Capped at 500 by default with meta.truncated. A format is optional — anything unrecognised is scanned as plain text rather than refused.
The server takes content and returns data — it reads no files and makes no network requests of its own. Published as numbers-le-mcp on npm and as io.github.nolindnaidoo/numbers-le in the MCP registry.
Most hosts read a JSON config. Add one entry:
{
"mcpServers": {
"numbers-le": {
"command": "npx",
"args": ["-y", "numbers-le-mcp"]
}
}
}-y skips the install prompt on first run. Pin a version if you would rather not track releases — numbers-le-mcp@2.5.0.
Prefer not to go through npx on every launch? Install it once and point at the binary instead:
npm install -g numbers-le-mcp{
"mcpServers": {
"numbers-le": { "command": "numbers-le-mcp" }
}
}It speaks MCP over stdio and needs no environment variables, no API key and no configuration of its own. To check it before wiring it into anything:
echo '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | npx -y numbers-le-mcpThat prints the tool list and exits — if you see extract_numbers, the server works.
Supported formats
Format | What counts as a number |
JSON | Number values anywhere in the structure. Quoted numbers ( |
YAML | Number values, including hex/octal int forms and exponents; multi-document ( |
TOML | Integer and float values (dates are not numbers) |
CSV | Cells whose entire content is numeric — |
INI | Values whose entire content is numeric (INI values are inherently strings) |
Environment | Values whose entire content is numeric ( |
The shared policy (one heuristics module, applied to every format): only finite numbers are extracted — NaN and inf/.inf never appear in results — and a coerced string must be numeric in full: plain decimals with optional sign, decimal point, and exponent.
Python, Rust, Go, Java, Kotlin, C#, C, C++, JavaScript, TypeScript, SQL and shell are read by a numeric-literal extractor that understands hex 0xFF, binary 0b1010, octal 0o755 and legacy 0755, digit separators 1_000_000 and 1'000, and suffixes 123n, 1.5f, 10u32, 100L — and that knows u32 is a type name, not the number 32. A dialect changes an answer: 0755 is 493 in C, C++, Go and Java, and 755 in Rust, Python 3, Kotlin and C#.
Every finding carries the notation it was written in — decimal, hex, binary, octal, scientific, bigint. For anything left over you can pick a format explicitly or fall back to a plain-text scan (no grammar: v1.2.3 reads as 1.2 and 0.3).
The CLI
The same extraction runs from a terminal or a shell pipeline: a Rust CLI
in crate/, sharing one corpus with the extension —
crate/fixtures/ — so the two can never read a
document differently. Install it from npm with npm install -g numbers-le
(or run it with npx numbers-le) or from crates.io with
cargo install numbers-le. Both install the same command and are held to
the same output byte for byte, apart from the words a parser uses to refuse
a broken document.
numbers-le . # every number in the tree; JSON on stdout when piped
numbers-le --values config/ # just the numbers, one per line
numbers-le mcp # the same extraction over MCP on stdio
# the point of the whole thing:
numbers-le --values --dedupe src/ | sort -n > after.txt
diff before.txt after.txt # what constants moved this releaseInstall it from npm (numbers-le) or from crates.io (numbers-le). Both install the same numbers-le command.
npm install -g numbers-le # or run it once: npx numbers-le
cargo install numbers-leSomebody has to verify that the rate in the code is the rate in the specification. A tax percentage, a retention window, a rounding boundary. In a regulated setting that check is a deliverable, and the person doing it usually has no checkout and never has the editor open.
Numbers are printed exactly as JavaScript prints them — 1e+21, not
Rust's 1000000000000000000000. This tool's whole output is numbers as
text, so the rendering is the contract, and the corpus pins it.
Exit codes follow grep — 0 numbers found, 1 none found, 2 the question was malformed.
Commands
Command | Description |
| Extract all numbers from the active document |
| Remove duplicate numbers from the results |
| Sort results numerically or by magnitude |
| Open Numbers-LE settings |
| Built-in documentation |
No command is bound to a key by default. Give any of them one under Keyboard Shortcuts in the editor.
Settings
Setting | Default | Description |
|
| Open results beside the current editor |
|
| Dedupe/Sort write to a new file instead of replacing the editor content |
|
| Show the line and column of each number |
|
| Also copy results to the clipboard (non-CSV) |
|
| Include the line and column in that copy |
|
| Deduplicate automatically during extraction |
|
| Sort automatically during extraction |
|
|
|
|
| Surface parse errors as notifications |
|
|
|
|
| Guardrails for very large files/outputs |
|
| Warn above this file size |
|
| Offer Open/Copy/Cancel above this result count |
|
| Confirm before opening this many column documents |
|
| Show the status bar item |
|
| Local-only event log (see Privacy) |
Languages
Twelve languages besides English:
German · Spanish · French · Indonesian · Italian · Japanese · Korean · Portuguese (Brazil) · Russian · Ukrainian · Vietnamese · Chinese (Simplified)
Both halves are covered — the manifest (command titles, setting names and descriptions) and everything shown while the extension runs (notifications, the status bar, quick-picks and prompts). The extension follows VS Code's display language, so it matches whatever the editor is already set to; no setting of its own.
Privacy & security
No network access. The extension never sends data anywhere. The
telemetryEnabledsetting only writes events to a local Output Channel you can inspect (Numbers-LE).The MCP server holds the same line. It takes content as an argument and returns data: no filesystem access, no network calls, no telemetry. Your agent already has file-read tools, so duplicating them inside the server would add a path-traversal surface for no capability.
check:mcp-bundlefails the build if the server ever imports something that could reach either.Error notifications redact home directories and credential-shaped fragments.
One rating prompt, at most twice. On the 3rd successful use the extension asks once whether you would rate it, and once more on the 20th if you chose Later or dismissed it. Don't Ask Again ends it. Setting
notificationsLeveltoimportantorsilentyourself turns it off. The counts are kept in VS Code's extension storage and nothing is sent anywhere; Rate opens the listing you installed from — the VS Code Marketplace or Open VSX — in your browser.
Documentation
What | Where |
What the tool is allowed to say — scope, output contract, refusals, non-goals | |
How the extension is built and held together — architecture, invariants, toolchain, release | |
How the CLI is built and held together | |
What changed | |
The tool's page, and the other fifteen |
Performance
Input | Size | Found | Time | Rate | Scan speed |
JSON metrics | 2.18 MB | 120,000 | 69.12 ms | 1,736,013/sec | 31.6 MB/s |
CSV data | 0.93 MB | 150,000 | 88.02 ms | 1,704,098/sec | 10.6 MB/s |
INI config | 0.42 MB | 30,000 | 23.01 ms | 1,303,868/sec | 18 MB/s |
Median of 7 runs after warmup, on Apple M5 Pro, 24 GB RAM, Node 24.3.0. Inputs are generated
by scripts/benchmark.ts rather than checked in, so the sizes above are
exactly what was measured. Reproduce with bun run benchmark.
These are machine-specific and are not asserted in CI — a benchmark that gates a build only tells you how busy the runner was.
Testing
Metric | Coverage |
Statements | 81.24% |
Branches | 72.41% |
Functions | 89.08% |
Lines | 82.72% |
526 test cases across 37 files, plus an integration suite that runs
in a real VS Code extension host and an end-to-end test that installs the
built .vsix into a clean profile.
Generated from a real run — coverage/coverage-summary.json and
coverage/test-results.json — by scripts/coverage-readme.js, and
bun run coverage:readme:check says whether it has drifted since. Reproduce
with bun run test:coverage, and the case count is the one vitest prints.
More from the LE family
Sixteen single-purpose tools for the work in front of every model. Each ships a Rust CLI and an MCP server. One page: letools.dev
Get it out
String-LE — Extract every string in a codebase, so a person can read them
Numbers-LE — Extract every hardcoded number in a codebase, so a person can check them
Units-LE — Extract every quantity with its unit, normalized, and refuse the ambiguous ones by name
Dates-LE — Extract every date and timestamp, and the exact instant each one resolves to
IDs-LE — Extract every UUID, ULID, NanoID, ObjectId and Snowflake, and decode the time inside
IPs-LE — Extract every IP address, CIDR block and MAC, normalized and classified by scope
URLs-LE — Extract every URL in a codebase, with its protocol and exact position
Paths-LE — Extract every file path in a codebase, and say whether it still points at anything
Colors-LE — Extract every color in a codebase, and say which ones are not in your palette
Check it
Regex-LE — Find every regex in a codebase, and report which can be driven into catastrophic backtracking
Versions-LE — Find where one dependency is constrained differently across a repository's manifests
i18n-LE — Identify the i18n library a project uses, then audit its catalogs by that library's rules
Scrape-LE — Check whether a page is scrapeable before the scraper is written, and say when it cannot tell
Guard it
Secrets-LE — Find hardcoded credentials in a codebase, and never print one into the report
EnvSync-LE — Compare the dotenv files in a tree, and say which keys are missing from which
Unicode-LE — Find the Unicode that hides meaning — bidi controls, invisibles, homoglyphs, mixed scripts
Each stands on its own: no shared crate, no published core. Where two of them agree, it is because the same answer was right twice.
Contact — nolindnaidoo.com · GitHub · LinkedIn
Also by nolindnaidoo
Rust — pixelcoords and pixelactions are one loop: pixelcoords answers where, pixelactions acts there. Their own tools, their own voice — not part of the LE family.
pixelcoords — Freeze your screen, mark regions, get pixel-exact coordinates and crops pixelcoords.dev · crates.io · docs.rs
pixelactions — Consume human-verified coordinates, perform the interaction, confirm it landed pixelactions.dev · crates.io · docs.rs
License
MIT © nolindnaidoo
This server cannot be deployed
Maintenance
Related MCP Connectors
300+ deterministic data utilities for AI agents: validate, normalize, parse, match, redact.
Deterministic AI agent microtools, no accounts/API keys. fetch_extract: 98% token cut. 47 tools.
Turn messy text into strict JSON schemas agents can trust (invoice, receipt, contact, resume).
500+ deterministic tools for AI agents: math, conversion, validation, hashing, encoding, date/time.
Related MCP Servers
- AlicenseAqualityAmaintenanceProvides LLMs with deterministic math tools for sandboxed expression evaluation, descriptive statistics, primality testing, prime factorization, GCD/LCM, quadratic solving, and base conversion. Every operation runs real code rather than relying on the model's arithmetic, with expression parsing restricted to a whitelist of operators, constants, and functions.2111 PyPIMIT
- AlicenseNot gradedqualityAmaintenanceEnables agents to pull every file path out of supplied document content — imports, asset references, and config values — and get each one classified as a file, relative, absolute, or URL with its line and column. It works purely on text passed in, reading no files and making no network requests, so codebases can be audited for references that still point at something.22 npm1MIT
- AlicenseNot gradedqualityAmaintenanceEnables AI agents to perform token-efficient local codebase search, decompile NuGet types, compress source code, and read or analyze PDF documents through targeted text, image, form, bookmark, attachment, and hyperlink extraction.MIT
- AlicenseAqualityAmaintenanceEncode, decode, and convert numbers across FP64, FP32, FP16, BF16, TF32, OCP FP8/FP6/FP4, MX, and integer formats, with bit-level breakdowns, rounding modes, and precision-loss analysis. Runs locally; no API key needed.57MIT