Quicky Wiki

Quicky Wiki reads your papers, notes and web pages, extracts the individual claims, and gives each one a confidence score with its source. It surfaces contradictions as they appear, lets claims strengthen or decay over time, and shows the result in a local dashboard with a knowledge graph and a chat that answers with citations. The same knowledge can be compiled to Obsidian-style markdown, slides or flashcards, and agents can use it through a built-in MCP server.

How it works

1. 01

   A language model extracts atomic claims from each source document and assigns each a confidence score.

2. 02

   Claims and an event log of every change in belief (created, reinforced, challenged, weakened, superseded, resolved) are stored in a local SQLite graph, and confidence changes cascade to dependent claims.

3. 03

   The graph is compiled into wiki pages, slides, flashcards and visualizations, and served through the dashboard and the MCP server.

What it does

• Every claim carries a confidence score, its provenance and the claims it depends on
• Contradictions and knowledge gaps are surfaced automatically when a new source is ingested
• Local dashboard with an interactive knowledge graph, claims, timeline, health view and Ask Wiki chat with citations
• Compiles to Obsidian-compatible markdown, Marp slides, Anki flashcards, a D3 graph and a timeline
• Built-in MCP server (stdio or HTTP) and an embeddable library export