# Codeknow

> Turn any codebase into a queryable knowledge graph with health scores, drift detection, impact analysis, and onboarding guides — all from AST parsing, no API keys required.

Codeknow is an open-source CLI tool by Alex Salsali that converts any codebase into a queryable knowledge graph using tree-sitter AST parsing across 25+ languages. It requires no API keys, no embeddings, and no vector stores for its core functionality, making it immediately usable after a single `pip install`. A live demo is available at codeknow.onrender.com where users can paste any GitHub repo URL and get architecture analysis instantly.

## What It Is

Codeknow sits in the code intelligence category, specifically targeting architectural understanding and developer productivity. It builds a NetworkX DiGraph where nodes represent symbols (functions, classes, modules) and edges represent relationships (imports, calls, inheritance). Every analysis command — from debt scoring to security tracing — operates directly on this graph without requiring an LLM. The tool works standalone or as a force multiplier for AI coding assistants including Claude Code, Cursor, Gemini CLI, OpenAI Codex, and VS Code Copilot Chat.

## Core Analysis Commands

Codeknow ships with a broad command surface covering the full software development lifecycle:

- **`codeknow debt`** — Composite architectural health score (0–100, letter grade) covering god-node concentration, cross-community coupling, import cycles, community cohesion, and dead code
- **`codeknow drift snapshot/compare`** — Save a named architecture baseline and detect what changed since
- **`codeknow impact "<file>"`** — Blast radius analysis showing what breaks if a file changes
- **`codeknow test-impact`** — Maps changed files to affected tests, pipeable directly to `pytest`
- **`codeknow security`** — Attack surface analysis via source-to-sink path tracing
- **`codeknow owners`** — Git-blame overlay for knowledge silos, bus factor, and CODEOWNERS generation
- **`codeknow refactor-plan "<target>"`** — Dependency-aware refactoring order with risk assessment
- **`codeknow onboard`** — Generates a guided codebase tour from graph topology
- **`codeknow simulate`** — Simulate removing, merging, or extracting nodes before making changes

All commands support `--json` for machine-readable output, enabling CI pipeline integration.

## Language and Integration Breadth

The tool supports 25+ languages through dedicated tree-sitter extractors: Python, JavaScript, TypeScript, Go, Rust, Java, C, C++, C#, Ruby, Kotlin, Scala, PHP, Swift, Lua, Zig, PowerShell, Elixir, Objective-C, Julia, Verilog, Fortran, Bash, Groovy, Apex, Dart, Pascal, OCaml, Common Lisp, Terraform (HCL), Robot Framework, SQL, JSON/config, and Markdown. Language detection is automatic.

For AI agent integration, Codeknow provides a one-command installer for each supported assistant (`codeknow install claude`, `codeknow install cursor`, etc.) and an optional MCP server (`pip install "codeknow[mcp]"`) that exposes `suggest_refactoring` and `impact_analysis` tools to any MCP-compatible client.

## Zero-LLM Architecture

The core pipeline — parse, build, analyze, simulate, discover, debt, drift, onboard, security, test-impact, owners, and refactor-plan — works entirely without any API key. LLM integration is optional and available via extras for OpenAI, Anthropic, and Ollama (local inference). This design means the tool is immediately useful in air-gapped or cost-sensitive environments. Optional extras also cover Neo4j and FalkorDB graph database export, PDF/Word/Excel/video document ingestion, PostgreSQL schema introspection, SVG export, and a file watcher for live rebuilds.

## Update: v1.1.2

The latest release is v1.1.2, published on September 2, 2026. The repository was created on August 31, 2026, and last updated on September 7, 2026, indicating active early development. The project is MIT-licensed and hosted on GitHub under the `asalsali/codeknow` repository.

## Features
- AST-based knowledge graph from source code (tree-sitter, 25+ languages)
- Architectural debt score (0-100, letter grade) with CI gating
- Drift detection with named baseline snapshots
- Impact analysis (blast radius) for file changes
- Test-impact mapping to run only affected tests
- Security attack surface analysis via source-to-sink path tracing
- Ownership analysis with git-blame overlay and CODEOWNERS generation
- Dependency-aware refactoring plan generation
- Onboarding guide generation from graph topology
- Simulation engine: remove, merge, or extract nodes before committing
- Interactive HTML visualization and TUI terminal navigator
- Live architecture dashboard at localhost:8787
- Export to Neo4j, Obsidian, SVG, GraphML, HTML, wiki
- MCP server for AI coding assistant integration
- One-command installer for Claude Code, Cursor, Gemini CLI, Codex, VS Code
- Zero-LLM default mode — no API keys required for core functionality
- Optional LLM integration (OpenAI, Anthropic, Ollama)
- Cross-repo global graph support
- Natural language query support (requires LLM)
- Incremental rebuild for changed files only
- All commands support --json for machine-readable output

## Integrations
Claude Code, Cursor, Gemini CLI, OpenAI Codex, VS Code Copilot Chat, Kilo Code, Google Antigravity, Kiro IDE/CLI, MCP (Model Context Protocol), Neo4j, FalkorDB, PostgreSQL, OpenAI API, Anthropic API, Ollama, GitHub Actions (CI), pytest, Obsidian

## Platforms
WEB, API, VSC_EXTENSION, CLI

## Pricing
Open Source

## Version
v1.1.2

## Links
- Website: https://github.com/asalsali/codeknow
- Documentation: https://github.com/asalsali/codeknow#readme
- Repository: https://github.com/asalsali/codeknow
- EveryDev.ai: https://www.everydev.ai/tools/codeknow