Complete CLI Command Reference for code-review-graph: build, update, visualize, wiki

code-review-graph provides four core CLI commands—build, update, visualize, and wiki—that manage the full lifecycle of a repository knowledge graph, from parsing source code to generating interactive visualizations and documentation.

The code-review-graph CLI, located in code_review_graph/cli.py, exposes a self-documented interface for creating, maintaining, and exploring code intelligence. This guide covers the complete code-review-graph command reference with exact syntax, flags, and source locations from the tirth8205/code-review-graph repository.

---

build Command: Parse Repository into Knowledge Graph

The code-review-graph build command performs the initial ingestion of your repository. It runs Tree-sitter parsers across all source files, extracts AST nodes (functions, classes, imports), and writes them to a SQLite database at .code-review-graph/graph.db.

Key behaviors:

• First run: Full parse of the entire codebase
• Subsequent runs: Incremental update (only changed files)
• Storage: SQLite database with nodes, edges, call graphs, and inheritance relationships

# Full initial build

code-review-graph build

# Skip computationally expensive flow analysis

code-review-graph build --skip-flows

# Skip post-processing (embeddings, community detection)

code-review-graph build --skip-postprocess

Internally, build dispatches to the MCP tool build_or_update_graph_tool as documented in docs/COMMANDS.md. The parsing layer uses language-specific Tree-sitter grammars; the graph layer stores entities and relationships in normalized tables.

---

update Command: Force Incremental Refresh

The code-review-graph update command is a convenience alias that triggers the same build_or_update_graph_tool with incremental semantics. Use this when you want to ensure the graph reflects recent edits without the cognitive overhead of choosing between full and incremental builds.

# Refresh graph after file changes

code-review-graph update

Both build and update support file-watching hooks for near-real-time synchronization (≈2.5 seconds for a 3,000-file project).

---

visualize Command: Generate Interactive Graphs and Exports

The code-review-graph visualize command transforms the SQLite knowledge graph into human-explorable formats. It supports multiple output targets and can launch a local server for immediate inspection.

Export formats:

• json – Raw graph payload for custom processing
• graphml – Import into Gephi, yEd, or Cytoscape
• svg – Static vector image
• obsidian – Obsidian vault-compatible Markdown
• cypher – Neo4j import statements

# Generate interactive D3.js HTML and open browser

code-review-graph visualize

# Export to GraphML for external tools

code-review-graph visualize --format graphml > mygraph.graphml

# Serve locally on custom port

code-review-graph visualize --serve --port 8765

# Export for Obsidian

code-review-graph visualize --format obsidian --output ./graph-vault

The visualization pipeline reads from graph.db, applies D3 force-directed layouts (for HTML output), or delegates to format-specific serializers in the MCP tool layer.

---

wiki Command: Browse Knowledge Base Locally or Export

The code-review-graph wiki command surfaces the structured graph data as a browsable wiki. This static site generation makes code relationships discoverable through hyperlinked pages rather than raw graph traversal.

Modes:

• --serve – Launch local HTTP server (default behavior)
• --format obsidian – Export Markdown vault for offline use
• --output <path> – Specify destination directory

# Open wiki in local server

code-review-graph wiki --serve

# Export as Obsidian-compatible Markdown

code-review-graph wiki --format obsidian --output ./wiki-export

The wiki generator renders entities (functions, classes, modules) as interlinked Markdown pages with automatic back-references and call hierarchy.

---

Architecture: How Commands Connect to Implementation

Understanding the CLI architecture clarifies where customizations and debugging should occur:

Layer	Location	Responsibility
CLI entry	code_review_graph/cli.py	Argument parsing, subcommand dispatch
MCP tools	Documented in docs/COMMANDS.md	Core operations (build_or_update_graph_tool, run_postprocess_tool, list_graph_stats_tool)
Graph storage	.code-review-graph/graph.db	SQLite persistence of nodes and edges
Parsers	Tree-sitter integrations	Per-language AST extraction
Visualizers	D3.js, format exporters	Human-readable output generation

All four commands (build, update, visualize, wiki) ultimately invoke MCP tool implementations that operate on the shared SQLite database.

---

Quick Reference: Command Comparison

Command	Primary Use	Typical Flags
build	Initial parse, full rebuild	--skip-flows, --skip-postprocess
update	Refresh after changes	(inherits build flags)
visualize	Explore graph visually	--format, --serve, --port, --output
wiki	Document relationships	--serve, --format obsidian, --output

---

Summary

• code-review-graph build creates or updates the SQLite knowledge graph via Tree-sitter parsing and MCP tools.
• code-review-graph update provides a semantic shortcut for incremental refreshes.
• code-review-graph visualize exports to multiple formats including interactive D3.js, GraphML, and Obsidian vaults.
• code-review-graph wiki generates browsable documentation from graph relationships.

The canonical documentation lives in docs/COMMANDS.md within the repository. Run any command with --help for flag details, or consult docs/USAGE.md for extended examples.

---

Frequently Asked Questions

Where is the official command documentation located?

The docs/COMMANDS.md file in the tirth8205/code-review-graph repository contains the definitive reference for all CLI commands, MCP tool signatures, and slash commands. This file is version-controlled and updated with each release.

What database format does code-review-graph use?

All graph data persists to a SQLite database at .code-review-graph/graph.db. This includes nodes (functions, classes, files), edges (calls, inheritance, imports), and computed metadata like embeddings and community assignments.

Can I export the knowledge graph for use in other tools?

Yes. The visualize command supports graphml for Gephi/yEd/Cytoscape, cypher for Neo4j, json for custom processing, and obsidian for Markdown-based tools. Use code-review-graph visualize --format <type> to specify your target format.

How do I browse the generated wiki offline?

Run code-review-graph wiki --format obsidian --output ./my-wiki to export a complete Markdown vault. This produces interlinked pages compatible with Obsidian, GitHub, or any static site generator without requiring a running server.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}