Introduction

Orstrum is a free scanner for TypeScript, JavaScript, and Python repositories. It walks your codebase with a full AST parser, resolves every import edge, and stores the result in a local SQLite database. That graph is then served over MCP to Claude Code — giving your agent structured, queryable codebase intelligence in a single tool call.

What it does

The scanner extracts four entity types: repositories, files (nodes), exports, and import edges. Barrel re-exports and path aliases from tsconfig.json are resolved to their actual targets — not approximated by text search.

Each file can optionally carry a purpose summary: an LLM-generated description of what the file does and why it exists. Summaries are computed once and cached. Push the graph to the cloud and your whole team benefits from summaries any member computed locally.

The three-step workflow

01

Scan — orstrum scan walks the repo AST and writes graph.db to .orstrum/.

02

Serve — orstrum serve starts a standard stdio MCP server over the local graph. Point Claude Code at it and queries are answered in milliseconds.

03

Push (optional) — orstrum push syncs the graph to your cloud workspace so teammates share the same index.

What never leaves your machine: raw source code. Orstrum stores file paths, export signatures, import edges, and purpose summaries only. The cloud graph is an architectural map — not a codebase copy.

Local vs. cloud

The local graph is always free. orstrum scan and orstrum serve work entirely offline — no account needed. Cloud sync (orstrum push) is the paid layer that makes the graph persistent and shared across your team.

Free local CLI

The CLI scanner and local MCP server are free to use — no account required. Your source code never leaves your machine; only the structured graph (paths, signatures, edges, and purpose summaries) is ever sent anywhere, and only when you run orstrum push. Cloud sync is the paid layer.