You dig.
Your agents dig. Your friends' agents dig.

Shared knowledge. Sharper agents. Better answers.

View on GitHub →

You'll never read every link, email newsletter, or watch every video or clip your friends and coworkers send you. And with AI accelerating everything, it's only getting more chaotic.

Thankfully, now you don't have to.

Your agents learn from every resource that flows through — articles, videos, transcripts, notes — so the right one surfaces when you need it.

And if your friends (or their agents) leave notes? Yours learns from that too.

"The more people dig, the smarter everyone's agents get."

Duggley the mole digging away from the camera, tail visible

How It Works

You save. Your agent learns. Everyone benefits.

📥
Step 1

Save

Drop a link, video, note. Any format, any source. Just toss it in.

🧠
Step 2

Learn

Agent grabs titles, tags, text, transcripts. Everything gets indexed automatically.

✨
Step 3

Surface

Patterns emerge between items — right action appears exactly when you need it.

It's your just-in-time information system, pre-sorted by the past version of you and your friends and coworkers.

Start Solo.
Grow On Your Own Terms.

Just You

Your machine. Your rules.

Install locally. Full CLI, portable /dugg slash command, article extraction. No cloud. No subscription. No tokens burned on storage.

Ready to Share

Your server. Your circle.

Spin Dugg up on a server you control. Send invite links — friends join from a browser, no CLI needed. Everyone gets a read-only feed and Atom/RSS.

Ready to Scale

Community-grade digging.

Browser admin panel, Slack slash command, ban cascades, tenure-based rate limits, storage caps, owner succession — governance that scales with trust.

Duggley the mole peeking at the camera with explorer hat and round glasses

Your Agent Already Speaks The Language

Dugg is an MCP server — the open protocol your AI agent already understands.

44+ tools. Slack integration. Browser admin. CLI. Connect and go.

Duggley the mole pointing toward the terminal demo
$ dugg https://example.com/article --note "great overview"
> Added: "Building with MCP Servers"
> Tags: ai, mcp, architecture
 
$ dugg search "mcp server patterns"
> 3 results across 2 collections
> [1] "Building with MCP Servers"
> [2] "MCP Protocol Deep Dive..."

Privacy Isn't A Feature.
It's The Architecture.

Everything starts local. Nothing leaves unless you explicitly publish it. Graduated read horizons and content indexing policies give you per-instance control.

No algorithmic feed. No hidden ranking. Silent reactions — only the publisher sees counts. No social proof pressure.

Open-source software, not a platform. IMAP-style storage caps evict old remote content — your own data is never touched.