What it does
SwarmVault takes material such as documents, code, transcripts, notes and URLs and compiles it into a folder of markdown pages plus a knowledge graph stored on disk. Your original sources are kept untouched in a raw folder, the generated wiki sits beside them, and a schema file lets you describe how the wiki should be organized. You can search and query the vault from the command line, open a local graph viewer, export the graph, and expose the vault to coding agents through a built-in MCP server or installable agent integrations. Search combines SQLite full-text with semantic embeddings, and graph edges are tagged as extracted, inferred or ambiguous so you can judge how far to trust them. No API keys are needed for the first run, because a built-in offline provider handles setup. For sharper extraction you can connect a local model through Ollama or a cloud provider such as OpenAI, Anthropic or Gemini. It installs as an npm command-line tool (Node 24 or newer) or as a desktop app for macOS, Windows and Linux.
Capabilities
- Compiles files, code, transcripts and URLs into a markdown wiki and knowledge graph
- Hybrid search that combines SQLite full-text with embeddings
- Contradiction detection and approval bundles to review changes before they land
- Built-in MCP server plus install targets for coding agents such as Claude Code and Codex
- Works offline with a built-in provider; local or cloud models are optional
- Local graph viewer and exports, including Neo4j
When to use it
- Building a personal or project knowledge base that stays on your own machine
- Giving a coding agent a searchable memory of a repository or document set
- Keeping research sources and notes organized as a linked wiki
