graphify: Turn Codebases into Queryable Knowledge Graphs

Overview

Graphify is a Python CLI and skill for AI coding assistants that maps project code and supporting files into a knowledge graph. It addresses the difficulty of navigating large codebases by making entities and their relationships queryable, rather than requiring repeated searches across files.

Code analysis uses local tree-sitter parsing, while semantic extraction for documents and media can use an assistant model or configured API backend. Graphify is a good fit when you want explainable structural exploration for a project, especially alongside an AI coding assistant. It is not a vector-search index.

Key Features

• Extracts code structure locally using tree-sitter, without an LLM call for code.
• Connects symbols across files, including imports, calls, and inheritance relationships.
• Labels relationships as extracted or inferred so users can distinguish source evidence from resolved links.
• Supports graph queries, shortest-path tracing, and explanations of individual concepts.
• Can include documentation, PDFs, images, video, audio, SQL schemas, and configuration files.
• Generates a graph visualization, a Markdown report, and a JSON graph for reuse.
• Integrates as a skill with multiple AI coding assistants and can expose graph queries through MCP.

Use Cases

• Developers onboarding to an unfamiliar repository can identify central concepts and trace how components connect.
• Engineers investigating a bug can follow a path between a user-facing feature and underlying implementation symbols.
• Teams can give AI coding assistants a structured project map to consult before searching or reading files.
• Maintainers can connect code with architecture notes, rationale, and documents when those files are included in extraction.

Project Facts

• Language: Python
• License: Apache-2.0
• Stars: 124k
• Forks: 11.9k
• Topics: ai-agents, antigravity, ast, claude-code, code-analysis, code-search, codex, cursor, developer-tools, gemini, graphrag, knowledge-graph, leiden, llm, mcp, openclaw, rag, skills, tree-sitter
• Archived: No

Getting Started

Install the package and register the assistant skill:

uv tool install graphifyy
graphify install

Then run /graphify . in a supported assistant. See the README for platform setup, optional extras, and command details.

Alternatives

• codegraph: CodeGraph focuses on local code structure, call paths, and change impact across languages, rather than graphs spanning code, documentation, and schemas.
• codebase-memory-mcp: codebase-memory-mcp centers on a persistent code-structure graph queried through an MCP server, rather than a broader graph of project files.
• GitNexus: GitNexus focuses on code dependencies and execution flows, exposing them through CLI and MCP rather than modeling documentation and schemas too.
• GitNexus: This GitNexus runs in the browser on GitHub repositories or ZIP files and presents an interactive graph, rather than building a local project-wide graph.

Considerations

• Python 3.10 or newer is required.
• Code extraction is local, but semantic extraction for documents and media uses an assistant model or configured backend. Headless extraction of those files may require provider credentials.
• Optional capabilities, including PDF, office, video, SQL, database, and MCP support, may require installing extras.
• The project distinguishes extracted relationships from inferred ones, so inferred links should be treated as resolved analysis rather than explicit source declarations.
• Graphify produces a graph for exploration, not a vector index; teams specifically seeking embedding-based semantic search may need a different or complementary tool.