# LLMGRAPH questions and answers

> Answers to common LLMGRAPH questions: which agents and languages it supports, what the labels mean, what it can't see, privacy and the trial.

Short answers to the questions people ask before they try LLMGRAPH. Each answer links to the page with the full detail.

## What is LLMGRAPH?

LLMGRAPH is a local MCP server and Claude Code plugin that tells your coding agent who calls a function, what breaks if it changes, and what it can't see. Every answer line carries a label for how far it is proven; [the docs](https://beta.llmgraph.in/docs.md) start from what you want to do.

## Is this the same as llmgraph.ai?

No. LLMGRAPH by Brainfoods Studio is not affiliated with llmgraph.ai, and it is separate from the llmgraph Python packages. [About LLMGRAPH](https://beta.llmgraph.in/about/) says who makes it and where it lives.

## Which coding agents does it work with?

Any agent that can use MCP servers can run it, and there are setup steps for seven. We ran Claude Code and Google Antigravity ourselves; the other five follow each vendor's docs and are not yet tested by us ([set up with your agent](https://beta.llmgraph.in/docs/start.md)).

## Does it work on Mac or Linux?

No. The beta runs on Windows 10 or 11 (x64) only, and on any other machine the launcher prints one line saying so and stops. [Set up with your agent](https://beta.llmgraph.in/docs/start.md) lists what you need.

## Which languages does it read?

Python and TypeScript/JavaScript. Other languages are not read, and [what it can't see](https://beta.llmgraph.in/docs/limits.md) lists the rest.

## How is it different from grep or my language server?

Use text search for a name that appears once and your language server for one exact definition. Use LLMGRAPH when names collide, or when you need to know which callers an edit missed and what can't be seen ([grep, language server or LLMGRAPH](https://beta.llmgraph.in/guides/grep-language-server-or-llmgraph.md)).

## What do the five labels mean?

Each answer line carries one of five labels, strongest first: certain, exactly_these, within_these, maybe_more and cannot_see. A label is never rounded up, and [the glossary](https://beta.llmgraph.in/docs/answers/what-the-labels-mean.md) explains each one.

## Can it be wrong?

In a blind test of 50 graded questions on three open-source repos, no line labelled certain, exactly_these or within_these was wrong. The same report lists where answers were misleading, and [real answers](https://beta.llmgraph.in/docs/real-answers.md) shows where it loses to a plain text search.

## What can't it see?

Calls built while the program runs, callers in code it could not read, and callers outside the project. Each one is counted and named in the answer instead of guessed ([where the answer is blind](https://beta.llmgraph.in/docs/questions/what_is_unknown.md)).

## Does my code leave my machine?

Your code is read and analysed on your machine, and we never receive code, paths, symbol or repo names, queries or answers. What does leave it is a licence check, plus usage counts and crash reports that you can switch off ([privacy](https://beta.llmgraph.in/privacy/)).

## Does it run my code?

The program never runs, imports or builds your code. The bundled type checkers may start your project's own interpreter to find installed packages ([terms](https://beta.llmgraph.in/terms/)).

## What does the stop check do in Claude Code?

When Claude Code ends a turn after editing, the plugin holds the agent once if callers it broke are certain to break, and names them by file and line. It never holds on weaker labels and can be switched off ([the stop check](https://beta.llmgraph.in/docs/stop-check.md)).

## How big a repo can it handle?

The beta plan allows up to 50,000 code lines per repository, in up to 3 repositories on 2 devices. [What it can't see](https://beta.llmgraph.in/docs/limits.md) lists the limits next to the languages.

## How much does it cost, and is there a trial?

There is one plan for individuals, $7 a month or $60 a year, with a 7-day trial that starts when you approve your first device. The beta is invite-only, and [pricing](https://beta.llmgraph.in/pricing/) has the terms.

## Does it work offline?

If the program cannot reach the licence service, it keeps working until the licence it already holds runs out, no later than 6 hours after the last renewal. After that, queries stop until the connection is back ([licence and devices](https://beta.llmgraph.in/docs/licence.md)).