Quick Start

This page walks through the full workflow from initialization to AI assistant integration.

1. Initialize Configuration

Run in your project root:

igraph init

This generates .igraph/config.json with default settings for the parser, LLM, vectorization, etc.

2. Provide Credentials

Option A: Global config (recommended — set once, shared across projects)

igraph config set apiKey sk-xxx

Option B: Environment variable (temporary override or CI)

export IGRAPH_API_KEY="sk-..."

Works without an API Key

Add --no-llm to skip LLM summaries and vectorization, using heuristic fallback. Retrieval automatically falls back to FTS5-only.

3. Build the Graph

igraph build

The first run performs a full build: parse → store → summarize → vectorize.

Common variants:

igraph build --no-llm    # Heuristic fallback (no API Key needed)
igraph build --dry-run   # Preview parse stats without writing to DB

4. Mount External Resources (Optional)

Mount PRD documents or DB schemas to establish cross-modal associations:

igraph mount prd docs/requirements.md
igraph mount db  schema/db.sql

5. Verify Retrieval

Query the graph with natural language to verify the build:

igraph query "where is user authentication implemented"
igraph query "JWT validation" --top-k 5 --json

6. Check Graph Status

igraph status

Outputs graph scale (files/nodes/edges), vector index status, mounted resources, etc.

7. Register with AI Assistants

Register the MCP Server with your installed AI assistants in one command:

igraph register

register auto-detects Claude Code and Cursor and writes the MCP config. You can also specify targets:

igraph register --target claude
igraph register --target cursor
igraph register --global        # Global config (shared across projects)

8. Manual MCP Server Start (Alternative)

If you prefer not to use register, start manually:

igraph serve

Then add the entry to your AI assistant's MCP config file manually. See MCP Integration.

Incremental Updates

After the first build, subsequent runs automatically do incremental updates (based on file SHA-256 diff):

igraph build    # Auto-detects changes, rebuilds affected files only

For a full rebuild:

igraph rebuild            # Clear and rebuild from scratch
igraph rebuild --no-llm   # Heuristic fallback rebuild

Next Steps

• Configuration — Full configuration reference
• Core Features — Deep dive into each capability
• CLI Reference — All commands and options