Graphify

Overview

Graphify builds a persistent, typed knowledge graph from code and related materials so AI coding assistants can query source-linked context instead of searching files. It works across repositories and can run as a hosted product or through the open-source engine on your own machine.

Key Features

• Builds a queryable knowledge graph from code, docs, SQL schemas, configs, and PDFs.
• Parses code locally with tree-sitter grammars, without uploading code or sending telemetry in the open-source tool.
• Serves graph context to assistants over MCP, including Claude Code, Cursor, Copilot, Codex, Gemini CLI, and others.
• Traverses a persistent cross-repo graph so answers can include file:line citations and evidence paths.
• Supports incremental updates by re-parsing changed files and can rebuild on commits with a git hook.
• Includes PR review triage and formal verification features, with coverage depending on workflow and language.

Use Cases

• Give coding assistants source-linked context for architecture questions and code navigation across repositories.
• Assess change impact and merge risk before merging, including CI, review, and conflict signals.
• Recover and share repository knowledge across teams without relearning the codebase by hand.
• Support audited or regulated work with graph paths and provenance for answers.

Getting Started

• Install the open-source CLI with uv tool install graphifyy.
• For Graphify Cloud, connect an MCP client to the Streamable HTTP MCP endpoint and complete OAuth sign-in.
• A Graphify Cloud setup requires a Graphify account, workspace access, and an indexed repository.
• Running graphify install registers the /graphify skill with the assistants it detects.

Deployment & Requirements

• Choose between the hosted Graphify Cloud service or the open-source engine on your own machine.
• The open-source engine can be served over stdio or HTTP on infrastructure you own.
• Enterprise can run the whole system on your own infrastructure, including air-gapped deployment.
• For fully offline use, point the semantic pass at a local model such as Ollama or any OpenAI-compatible server like vLLM.

Before You Adopt

• License: Apache-2.0. Review its terms before using, modifying, or distributing the project.
• The open-source tool is designed not to send code anywhere; the hosted product is an optional alternative.
• Local parsing keeps code on the machine, but the non-code semantic pass may leave the machine if pointed at a hosted model.
• The hosted product keeps the graph in the cloud on connected repositories, so using Cloud involves a hosted service and connected repositories.
• The enterprise layer is described as early access, so teams should account for that maturity status.
• Formal verification coverage is not universal and depends on the workflow and language.