Auto-Captured URLs

How Coppermind automatically detects and tags URLs you share in conversation, making links searchable and preventing the "I already gave you that link" frustration.

The Problem#

During client work, you share URLs constantly -- Google Drive folders, campaign dashboards, competitor websites, reference articles, shared docs. In a normal conversation, those links are lost the moment the session ends. The next time you need that link, you have to dig through chat history or ask the client again.

Coppermind solves this by automatically detecting URLs in any memory you store and tagging them for easy retrieval.

How It Works#

Whenever you store something -- a full memory, a quick note, or just mention it in conversation -- Coppermind scans the content for links. If it finds any, it automatically:

1. Adds url and shared_resource tags to the memory
2. Returns the captured URLs in the response so you can confirm they were picked up

This happens silently. You do not need to do anything special -- just mention a link naturally.

"Remember: Acme's brand guidelines are at https://drive.google.com/drive/folders/abc123"

Coppermind stores the memory as usual, detects the URL, and adds the tags. The response includes:

{
  "memory_id": "...",
  "summary": "Acme's brand guidelines are at https://drive.google.com/dri...",
  "urls_captured": 1,
  "urls": ["https://drive.google.com/drive/folders/abc123"]
}

Finding URLs Later#

Because URLs are tagged, you can search for them:

"What links have I shared for Acme?"

Or search by what the link was about:

"Where are Acme's brand guidelines?"

Both approaches work. The first uses tag-based filtering. The second uses semantic search -- the memory content around the URL provides the context Coppermind needs to match it.

What Gets Captured#

Real-World Examples#

Sharing a Client Resource#

You: "Store this: Acme's Q2 campaign tracker is at
      https://docs.google.com/spreadsheets/d/1abc123/edit"

Coppermind: Stored. 1 URL captured.

Three weeks later:

You: "Where's Acme's campaign tracker?"

Coppermind: "Acme's Q2 campaign tracker is at
https://docs.google.com/spreadsheets/d/1abc123/edit
(stored March 5, tagged: url, shared_resource)"

Multiple URLs in One Memory#

You: "Remember: Acme's competitor analysis sources are
      https://www.competitor-a.com and https://www.competitor-b.com/pricing"

Coppermind: Stored. 2 URLs captured.

The response shows the first 5 URLs found. All are tagged on the memory.

URLs During Transcript Ingestion#

When you ingest a meeting transcript that contains URLs (e.g., someone shared a screen with a link visible, or said "check out example.com/demo"), those URLs are preserved in the extracted memories and tagged the same way.

What Is NOT Captured#

• URLs are not crawled or summarized. Coppermind stores the URL as a reference, not the page content. It does not visit the link.
• URL shorteners are stored as-is. A bit.ly link stays as bit.ly -- Coppermind does not expand it.
• No bookmark management. This is not a bookmark tool. URLs are just another type of knowledge attached to memories. They follow the same search, decay, and handoff rules as any other memory.
• No validity checking. Coppermind does not check if the URL still works. Dead links stay in memory until you hide or replace them.

Deduplication#

If you already tagged a memory as a link yourself, Coppermind leaves your tag alone -- auto-capture only fills in when there's no link tag yet.

Saved links also benefit from the usual smart dedup: store the same link with very similar context twice and Coppermind updates the existing note instead of creating a duplicate.

Key Details#

• Automatic and silent. No configuration needed. URLs are detected and tagged on every memory store.
• Tags added: url and shared_resource. These are appended to any existing tags on the memory.
• First 5 URLs returned. The response shows up to 5 detected URLs. If the content has more, all are tagged but only the first 5 appear in the response.
• Code blocks are excluded. Fenced code blocks (triple backticks) and inline code (single backticks) are stripped before URL scanning to avoid capturing example URLs from technical content.
• Works across all clients. URL capture is per-memory, scoped to the active client mind like everything else.