# Llmgraph

> Llmgraph is an MCP server for developer tools. Call-graph answers for coding agents that say what is certain and what they cannot see. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Free trial
- **Platforms:** Windows, CLI
- **Website:** https://beta.llmgraph.in
- **Repository:** https://github.com/brainfoods-studio/llmgraph-agent-plugin
- **MCP Registry name:** io.github.brainfoods-studio/llmgraph
- **Categories:** developer tools
- **Listing updated:** 2026-11-14

## About

Searching code by text mixes look-alike functions together, so an agent can edit the wrong one or miss a caller. Llmgraph answers questions such as who_calls and check_my_changes with labelled results. A call is marked certain only when the type checker has settled that exact target. Other labels, such as within_these, record a known set of candidates, and same-name calls that point elsewhere are listed as ruled out with the reason.

Its distinguishing habit is naming blind spots: dynamic calls, code outside the repository and unread files are counted rather than hidden, and any line resting on an assumption is tagged. The README's example is zod's two copies of one helper, which text search mixes together. The README claims that in a blind test of 50 questions on three open-source repositories no confident answer was wrong.

It runs locally as an npm package and Claude Code plugin, for Python and TypeScript or JavaScript on Windows x64. The beta plan limits repositories to 50,000 lines. A 7-day trial needs no card, and the site says the price is locked for life for beta subscribers. Only licence checks, usage counts and crash reports leave the machine, and the counts and reports can be switched off.

## Features

- **Who calls this:** Lists callers of a function with a certainty label on each.
- **Change impact check:** Names the calls that break after a signature change and flags ones it cannot confirm.
- **Look-alike filtering:** Rules out same-name calls and states why.
- **Counted blind spots:** Reports dynamic calls and unread code instead of hiding them.
- **Stated assumptions:** Tags any line that depends on an assumption such as repo-only subclasses.
- **Local analysis:** Runs on your machine; the code is meant to stay there.

## Good for

- Find every caller before changing a function signature
- Check which callers a recent edit missed
- Separate two functions that share a name
- Know which parts of a call graph are unverified
- Reduce agent guessing during refactors

## Install

### Claude Code (Terminal)

```sh
claude mcp add llmgraph -- npx -y llmgraph-mcp
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "llmgraph": {
      "command": "npx",
      "args": [
        "-y",
        "llmgraph-mcp"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "llmgraph": {
      "command": "npx",
      "args": [
        "-y",
        "llmgraph-mcp"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "llmgraph": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "llmgraph-mcp"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "llmgraph": {
      "command": "npx",
      "args": [
        "-y",
        "llmgraph-mcp"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "llmgraph": {
      "command": "npx",
      "args": [
        "-y",
        "llmgraph-mcp"
      ]
    }
  }
}
```

## Tools

- `who_calls`: Lists the callers of a function with certainty labels.
- `check_my_changes`: Reports which calls break after files changed since the last call.

## FAQ

### Which languages does it support?

Python and TypeScript or JavaScript only, on Windows x64.

### Does my code leave my machine?

The site says it is built so code stays local. A licence check is sent, plus daily usage counts and crash reports unless switched off.

### Is there a size limit?

Yes. The beta plan indexes repositories up to 50,000 code lines.

### Is there a trial?

Yes, 7 days with no card, starting when you approve your first device.

### Is Llmgraph free?

Llmgraph is a paid product with a free trial.

### How do I add Llmgraph to Claude Code?

Run this command in your terminal: claude mcp add llmgraph -- npx -y llmgraph-mcp

### What are the best alternatives to Llmgraph?

The closest alternatives to Llmgraph are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Concord MCP](https://ai-powered.com/mcp/concord-mcp): Concord MCP is a freemium MCP server for developer tools.

---
Source: https://ai-powered.com/mcp/llmgraph
