FAQ

Frequently asked questions about Lore.

What's the difference between Lore and a memory system?

Memory systems store processed facts without attribution. Lore preserves original documents so you can cite exactly what was said, by whom, and when. When an AI tool uses Lore, it can say "In the Jan 15 interview, Sarah said..." rather than just "Users want faster exports."

What does sync cost?

Discovery (Phase 1) is free — Lore computes content hashes and checks which files are new without any LLM calls. Only new files trigger LLM processing (Phase 2) for metadata extraction and embedding generation. Re-syncing existing files costs nothing.

Can multiple people share a knowledge base?

Not yet. Each user's data is isolated via Postgres Row Level Security. Team sharing is planned for a future release.

What file formats are supported?

Markdown, JSONL, JSON, plain text, CSV, HTML, XML, PDF, and images (JPEG, PNG, GIF, WebP). Claude extracts metadata automatically during sync — including EXIF data from images and transcripts from structured JSON.

When you add a sync source with lore sync add, the default glob is **/* (all files). Use --glob to restrict to specific types if needed. The ingest MCP tool accepts any text content directly — just tell your AI "save this to lore."

Do I need to set up any infrastructure?

No. The backend (Supabase with pgvector) is fully hosted. Just install, log in, and bring your own OpenAI and Anthropic API keys.

How does multi-machine sync work?

Run lore setup on each machine with the same email. Your data repo URL is saved to your account, so new machines discover and clone it automatically. Lore deduplicates by content hash (SHA256), so nothing gets indexed twice even if the same document is synced from multiple machines.

How much does it cost to use?

Lore itself is free. You pay for your own API keys:

• OpenAI — Embedding generation (text-embedding-3-small). Very cheap, ~$0.02 per 1M tokens. Get a key at platform.openai.com/api-keys.
• Anthropic — Metadata extraction during sync, and the research tool. Costs vary by usage. Get a key at console.anthropic.com/settings/keys.

Is my data private?

Yes. Documents are stored in your local data directory and a Supabase database scoped to your user ID. Row Level Security ensures only you can access your data. API keys are stored locally in ~/.config/lore/config.json with restricted file permissions.

Can I use Lore without MCP?

Yes. The CLI works standalone:

• lore search "query" — search from the terminal
• lore browse — interactive TUI browser
• lore research "question" — AI-powered research

MCP integration is optional but recommended for AI-assisted workflows.

What AI tools work with Lore?

Any MCP-compatible client: Claude Code, Claude Desktop, Cursor, Windsurf, Gemini CLI, Codex CLI, and more. See the MCP Setup guide for platform-specific configuration.