Skip to main content

Quickstart

Install KMGraph, initialize a knowledge graph, and capture your first lesson — in under 5 minutes.

Step 1 — Install​

claude --plugin-dir /path/to/knowledge-graph

Verify the plugin loaded: type /km and the autocomplete menu should show kmgraph commands.

Step 2 — Initialize​

KMGraph init demo — wizard creates knowledge graph structure 
/kmgraph:init

The wizard asks for:

  • Project name
  • Git tracking — enable for automatic branch/commit metadata on every capture
  • KG type — project-local (default) stores in the project; personal stores at ~/.kmgraph/ and is shared across all projects

Init also scaffolds knowledge/me.md (your identity and working style, gitignored) and knowledge/rules.md (project conventions, committed). These give any AI platform consistent context without platform-specific rewrites. See Portable AI Identity.

Step 3 — Capture your first lesson​

KMGraph capture-lesson demo — structured lesson saved with git metadata 
/kmgraph:capture-lesson

Describe what you just learned or solved. The command structures it into a markdown file with categories, tags, and git metadata.

Step 4 — Recall it​

KMGraph recall demo — full-text search returns matching lessons 
/kmgraph:recall "what you captured"

Full-text search across all captured lessons, decisions, and patterns.

Step 5 — Check status​

KMGraph status demo — shows entry counts and last capture time 
/kmgraph:status

Shows the active knowledge graph, entry count, and recent captures.

The capture pipeline​

Use /kmgraph:sync-all at the end of a session to run all four steps in one command.

Next steps​

  • Tutorials — Hand-held walkthroughs for common capture scenarios
  • How-to Guides — Recipes for specific tasks (ADRs, meta-issues, sharing)
  • Concepts — How the knowledge graph is structured and why
  • Cheat Sheet — All commands at a glance