v0.1.0 Alpha — Now on NPM

Your AI was reading
the whole library.

Nexus-Graph gives your AI a surgical map of your codebase. Cut your token bill by up to 95% — by sending only the 5% of code that actually matters.

Get Started Free View on NPM
1,200+ files
15k+ symbols
42k+ edges
 97% Token Savings
The Flow

Three steps to smarter AI.

No PhD required. Just run the command and watch your token bill drop.

01

Scan

Nexus reads your entire codebase and maps every function, class, and dependency into a local database.

02

Map

A high-speed symbol graph is built locally. It knows which code talks to which — instantly.

03

Optimize

Your AI only sees the 5% of code it actually needs. No noise. No wasted tokens. Just precision.

Architecture

How the pieces connect.

Your Codebase
Nexus Indexer
Symbol Graph
MCP Server
Your AI
Plain English

What does it actually do?

No jargon. Here's the real story.

📚

Without Nexus-Graph

Imagine asking a librarian one simple question, and they respond by handing you the entire library. That's what AI coding assistants do without smart context. They read every file, every line — even the parts that have nothing to do with your question. You pay for all of it.

🎯

With Nexus-Graph

Now the librarian has a perfect index. They walk straight to the exact page you need. Nexus-Graph builds that index for your code — so AI only reads what matters. Faster answers, dramatically lower costs, and significantly less confusion.

Save on AI Bills

Most developers spend 70–95% of their tokens on irrelevant code context. Nexus eliminates that waste entirely.

More Accurate Answers

When an AI sees less noise, it thinks more clearly. Surgical context means fewer hallucinations and better code suggestions.

Your Code Stays Local

Nexus-Graph never uploads your source code anywhere. The graph lives on your machine. Only the relevant snippets travel to your AI of choice.

Works at Any Scale

Whether you have 10 files or 10,000, Nexus builds the same precise map. It was designed for the codebases that break every other tool.

FAQ

Frequently Asked Questions

Does Nexus-Graph upload my code to the internet?
No. Nexus-Graph runs 100% locally on your machine. It builds a symbol graph stored in a local SQLite database. Only the small, relevant code snippets it retrieves are passed to your AI assistant — and that happens through your own API key, not our servers.
Do I need to be a developer to use this?
You need Node.js installed and a basic comfort with the terminal. If you can run npm install, you can use Nexus-Graph. The commands are three lines. Once it's running, your AI assistant handles everything automatically.
Which AI tools does it work with?
Nexus-Graph works as an MCP (Model Context Protocol) server, which means it integrates with Claude Code, Cursor, and any AI assistant that supports the MCP standard. If your assistant can read from a local server, it works with Nexus.
How much can I actually save on my API bill?
It depends on your codebase size and workflow. On a typical project with 500+ files, users report sending 5–10% of the code they used to send per query. That directly translates to 72–95% fewer tokens charged per AI interaction. Bigger codebases see bigger savings.
What programming languages are supported?
Nexus-Graph currently supports deep indexing for TypeScript, JavaScript, and Python using AST (Abstract Syntax Tree) parsing. Support for additional languages is on the roadmap. Even unsupported file types benefit from the graph structure around them.
Is it free? What's the catch?
The core tool is fully open source under the MIT license — free for personal and commercial use, forever. There is no catch. The Enterprise plan (coming soon) will offer team sharing, cloud sync, and dedicated support for organizations that need it.
Plans

Open source at the core.

Start free. Scale when you're ready.

Open Source
Free

For individual developers who want smarter AI without the bill.

  • Local Graph Indexing
  • MCP Server Support
  • Works with Cursor & Claude
  • 100% Private — Runs Locally
  • MIT License
View on GitHub

Up and running in 60 seconds.

Install, index, and connect. That's it.

# 1 · Install the CLI globally npm install -g @costline/nexus-graph
# 2 · Index your project (run inside any folder) nexus-graph index
# 3 · Start the MCP server and connect your AI nexus-graph server
Enterprise

Ready to scale?

Tell us about your team and we'll get back to you within 24 hours.