Flywheel FAQs

Question 1

What kinds of intelligence can Flywheel extract from my knowledge base?

Accepted Answer

Flywheel offers three core query types: Graph queries (e.g., backlinks, connection paths, orphan notes), Schema queries (on note frontmatter without file reads), and Temporal queries (e.g., recent activity, stale notes). This unlocks deep insights into your knowledge structure.

Question 2

How does Flywheel achieve significant token savings for AI agents?

Accepted Answer

Instead of an AI agent burning thousands of tokens by reading your entire vault for every query, Flywheel allows it to query a compact, optimized index using around 50 tokens. This enables agents to perform hundreds of queries during long-running tasks without hitting context window limits or excessive costs.

Question 3

Is Flywheel secure and compatible with existing knowledge management setups?

Accepted Answer

Yes, Flywheel is Git-native and built for plain text markdown, ensuring future-proof privacy by keeping your files local and eliminating vendor lock-in. It's editor-agnostic, compatible with Obsidian, VSCode, and other markdown tools, seamlessly integrating into your workflow with an MCP client like Claude Code.

Question 4

What is Flywheel and its main purpose?

Accepted Answer

Flywheel is a graph intelligence server that indexes Obsidian vaults (or any markdown folder) to provide semantic search capabilities for AI agents. Its primary goal is to drastically reduce AI agent token consumption—up to 100x—by querying an index instead of feeding entire vault contents.

Flywheel

Flywheel

Key Features

Use Cases

Key Features

Use Cases