The Knowledge Base is live. Every link, quote, chat, or code snippet you add gets read, synthesised, and integrated into a structured store that compounds over time. Your agents always start informed — not from zero.
How it works
When you add a source, xysq:
- Ingests the raw content (unchanged — you can delete it any time)
- Indexes it in the background (usually seconds)
- Makes it available to your agents via
memory_recall
Sources are linked to your identity, so any connected agent — Claude, Cursor, or others — can draw on the same knowledge base.
Adding sources
You can add to your knowledge base via your AI agent using the MCP connection.
Via your agent (MCP)
Once you have Claude or Cursor connected, xysq works in the background. As you have conversations, share links, drop in transcripts, or discuss code — xysq automatically identifies what’s worth keeping and adds it to your knowledge base. You don’t need to tell it to save anything.
Just work naturally:
Share a link
Here's an article I found on database indexing: https://use-the-index-luke.com
Paste a chat transcript
Here's a conversation I had earlier about the auth middleware — can you pick up from here?
[paste transcript]
Discuss a code snippet
Here's the middleware I've been working on (src/auth/middleware.py, lines 45–80):
[paste code]
xysq captures what matters from each interaction and builds it into your knowledge base automatically.
Documents (PDFs, uploaded files) are not yet supported. Document ingestion — via the dashboard and MCP clients — is coming soon.In the meantime, paste the text content as a quote instead.
Source types
| Type | What it is | How to add |
|---|
link | A URL to fetch and index | Ask your agent |
quote | Pasted text, article excerpts, notes | Ask your agent |
chat | A conversation transcript | Ask your agent |
code | A code snippet (language auto-detected) | Ask your agent with a file path and line range |
document | A file (PDF, etc.) | Coming soon |
Recalling knowledge
Once indexed, knowledge surfaces automatically when your agent answers questions. You can also ask explicitly:
What do you know about database indexing?
What have I saved about auth middleware?
Your agent will draw on both memories (episodic — what happened) and knowledge (structural — what it means). The knowledge base is the source of truth both layers build from.
Deleting sources
You own your knowledge base. Delete any source at any time from the dashboard — it will be removed from future recall results.
Delete the database indexing article I saved earlier.
Or delete directly from the Knowledge Base page in app.xysq.ai.