How crowded is the AI-agent niche on Gumroad? A live density tracker

Everyone says "the AI-agent space is exploding." That is a vibe, not a number. So I count it: every week I run a handful of Gumroad searches and log how many products come back. Below is the running series, what it means, and how to run it yourself in 30 seconds.

The current numbers

Each cell is the "total" result count Gumroad returns for that search term, captured with a single curl per query — no login, no scraping. Most recent read: 26 July 2026 (next scheduled read ~26 July, weekly cadence).

"—" means I had not started logging that term yet on that date.

How to read it — supply is not demand

The single most important thing on this page: this is a supply count, not a demand signal. A search total tells you how many products exist for a term — not whether any of them sell. So use it for exactly one thing: finding niches that are thin but moving, then validating demand separately.

• Saturated (notion ~19,737, chatgpt prompts ~4,731): tens of thousands of competitors. You can win here, but only with a real distribution edge.
• Thin and very thin (claude skills 669, mcp server 157, openclaw 157): few competitors, so a clear, useful product can stand out — if people are actually buying. That is the part you still have to check.

The one trend that matters

Raw counts wobble. What I watch is the relative week-over-week move. On the latest read (19→26 Jul) the saturated notion term actually ticked down while the thin AI-agent terms rose — the cleanest niche signal yet, because it is clearly not just broad catalogue growth:

Nine consecutive rises turn "claude skills is growing" from a two-point guess into a firmly confirmed trend. And this read is the cleanest yet: the saturated notion template term actually fell while the thin AI-agent terms climbed, so this is genuine niche movement, not the whole catalogue drifting up. Meanwhile last month's openclaw +10.79% spike has now sat flat for two reads running — confirming it was a one-read blip I was right not to chase. That discipline — catching early signals and filtering out the false ones — is exactly why I publish a full report on this niche instead of just a snapshot.

The exact method — reproduce it in 30 seconds

No special access. Gumroad's public product search returns a "total" field. One curl per term, one polite second between calls:

for q in "openclaw" "claude skills" "mcp server" "n8n template" \
         "ai agent prompts" "ai agent" "chatgpt prompts" "notion template"; do
  enc=$(printf '%s' "$q" | sed 's/ /%20/g')
  n=$(curl -s "https://gumroad.com/products/search?query=$enc" \
      | grep -o '"total":[0-9]*' | head -1 | grep -o '[0-9]*')
  printf '%-20s => %s\n' "$q" "$n"
  sleep 1
done

Run it weekly, append a column, and you have your own density series for any niche you care about — not just these terms.

Honest caveats

• Supply, not demand. Said twice on purpose. A growing count can mean opportunity or a gold rush of products that don't sell. Pair it with a sales sample.
• Search totals are noisy. Gumroad's index shifts; trust the trend across several reads, not any single number.
• Terms overlap. "ai agent" partly contains the narrower terms, so don't add the rows together.
• I have skin in this. I sell products in this niche, so treat the data as a starting point and verify it yourself — the method above lets you do exactly that.

Related reading: Claude Skills explained · OpenClaw skills, explained · n8n AI agent workflows.