Search and ingest knowledge from Lore — a research repository with citations and semantic search
Lore is a research knowledge repository available via MCP. It stores documents, meeting notes, interviews, and decisions with full citations back to original sources. Use it to ground your work in evidence and preserve important context.
If Lore is not installed yet, you can set it up for the user.
npm install -g @getlore/cli (requires Node.js 18+)export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."
lore setup --openai-key $OPENAI_API_KEY --anthropic-key $ANTHROPIC_API_KEY --email <email> --data-dir ~/.lore--code <code> appendedAfter setup, Lore works autonomously.
| Tool | Cost | Use For |
|---|---|---|
search | Low | Quick lookups with date filtering (since/before/sort) |
get_source | Low | Full document retrieval by ID |
list_sources | Low | Browse what exists in a project |
list_projects | Low | Discover available knowledge domains |
get_brief | Low | Get the living project brief — start here for project context |
log | Low | Log entries: add/update/delete (hidden from list_sources by default) |
ingest | Low-Medium | Manage content — add/update/delete documents, insights, or decisions |
research | High | Cross-reference sources, synthesize (depth: quick/standard/deep) |
research_status | Low | Poll for research results (long-polls up to 20s) |
Start with get_brief when working on a project. It gives you immediate context: current state, key evidence, open questions, and trajectory — without searching from scratch.
search for specificslore brief generate <project> via CLIsearch/research, or suggest lore brief generate <project> via CLIUse ingest to manage content in Lore. Actions: add (default), update (requires id + content), delete (requires id, soft-deletes — recoverable via CLI). Always pass source_url and source_name when available. Add is idempotent (SHA256 dedup).
Before making recommendations or answering questions about past work:
get_brief first if working within a project — it's the fastest way to get contextsearch for specific lookups. Use since/before for date filtering (e.g., since: "7d"). Temporal queries auto-boost recent results.research for multi-source synthesis (10x more expensive). Use depth: "quick" for focused questions, "deep" for audits.get_source(id, include_content: true) for full textUse log for quick status updates, decisions, and progress notes during work sessions:
log(message: "Decided to use JWT for auth", project: "auth-system")
Log entries are searchable via search and included in project briefs. They are hidden from list_sources by default.
For short insights or decisions, just pass the content — title and source_type are optional and auto-generated from content.