GumroadDNA MCP¶

A creator storefront, operated from the conversation. GumroadDNA puts the Gumroad API v2 behind 44 tools so an assistant can run a Gumroad store end to end: create a product (file, cover, thumbnail, publish), manage images and stock, read and refund sales, resend receipts, manage license keys, offer codes, custom fields and webhooks, and pull a today, week and month revenue snapshot. It also reads the public Gumroad storefront for any keyword - how many products, the price spread, top sellers - so a niche can be sized before entering it.

Server type: Local (stdio, Python 3.10+; optional HTTP via PORT)
Auth: Your own Gumroad access token (GUMROAD_ACCESS_TOKEN)
Install: git clone the repository, pip install -r requirements.txt
License: MIT
Built by: VibeDNA

Why This Matters for Operators¶

Gumroad is where a lot of small software, ebooks and courses get sold, and its dashboard is where creators lose evenings. This server moves the routine operations into the agent: restocking a product file, refunding an order, resending a receipt, checking which license keys are still valid, creating an offer code for a launch. The niche scan adds the market side: point it at a keyword and it reads the public storefront for product counts, price spread and top sellers by rating count, with no token needed at all.

Writes act on your live store, so the workflow stays deliberate. The vendor's guidance is that the AI client asks before calling a write tool unless you have allowed it; destructive tools like refund_sale and archive_product are labeled as such. For a store measured in evenings, that is the right shape: everything reachable, nothing silent.

Tools¶

44 tools across eleven areas:

Area	Tools
Market research	`scan_gumroad_niche` (public storefront, no token needed)
Products	`create_product`, `launch_product`, `update_product`, `publish_product`, `archive_product`, `list_products`, `download_product_files`
Files and images	`add_product_file`, `set_product_cover`, `add_product_image`, `list_product_images`, `remove_product_image`, `reorder_product_images`, `download_product_images`
Sales	`get_revenue_snapshot`, `list_sales`, `get_sale`, `refund_sale`, `resend_receipt`, `mark_as_shipped`
Subscribers	`list_subscribers`, `get_subscriber`
License keys	`verify_license`, `enable_license`, `disable_license`, `decrement_license_uses`
Discounts and checkout	`list_offer_codes`, `create_offer_code`, `delete_offer_code`, `list_custom_fields`, `create_custom_field`, `delete_custom_field`
Stock	`get_stock_status`, `update_stock`, `bump_stock`
Webhooks	`list_webhooks`, `create_webhook`, `delete_webhook`
Extras	`scan_unlisted_assets`, `sync_buyers_to_resend`
Account and meta	`get_account`, `vibedna_info`, `check_for_update`

Installation¶

Get an access token from Gumroad under Settings, Advanced, Applications. Then:

git clone https://github.com/marstudio360/gumroad-dna-mcp
cd gumroad-dna-mcp
pip install -r requirements.txt
claude mcp add gumroad-dna --scope user \
  -e GUMROAD_ACCESS_TOKEN=your_token \
  -- python /absolute/path/to/gumroad-dna-mcp/server.py

JSON clients use the same shape:

{
  "mcpServers": {
    "gumroad-dna": {
      "command": "python",
      "args": ["/absolute/path/to/gumroad-dna-mcp/server.py"],
      "env": { "GUMROAD_ACCESS_TOKEN": "your_token" }
    }
  }
}

The token can live in the MCP entry or in a .env file next to server.py. The repository ships an INSTALL.md written to be pasted into an AI chat and installed by the assistant itself.

Business Relevance¶

Indie sellers and creators restock files, launch products and answer revenue questions without opening the Gumroad dashboard.
Small teams handle refunds, receipts and shipped orders from wherever the conversation happens.
Software sellers check, enable and disable license keys from the same chat that answers support.
Anyone validating a niche reads the public storefront for counts, price spread and top sellers before building.

Integration with CorpusIQ¶

The store is one revenue surface among several. GumroadDNA answers what sold on Gumroad and what the store looks like; CorpusIQ reads the surrounding picture from Stripe, GA4, the ad platforms and 40+ connectors. An operator can compare how a launch performed on the store against where the traffic came from, both pulled read-only at answer time - and unlike the read-only connectors, decide for themselves when to let the agent act on the store through the Gumroad tools.

Limitations¶

Local install: it runs on your machine over stdio; there is no hosted endpoint.
Write tools act on the live store immediately; destructive ones are labeled and should keep the client's confirmation step on.
A Gumroad access token is required for everything except the public niche scan; token custody is yours.
Scoped to Gumroad only; other storefronts need their own connectors.
The niche scan reads the public storefront surface, not Gumroad's internal analytics.

FAQ¶

Do I need a paid Gumroad plan?¶

No. The server works with a standard Gumroad access token; the same API surfaces apply across plans, and the niche scan needs only public storefront data.

Can an agent create and publish a product end to end?¶

Yes: create_product plus file, cover and thumbnail tools, then publish_product. The client asks before writes unless you have allowed them.

What is the niche scan?¶

scan_gumroad_niche reads the public Gumroad storefront for a keyword and returns how many products exist, the price spread, and top sellers by rating count - useful before committing to a product idea.

Where does it run?¶

Locally over stdio with Python 3.10+, from the MIT-licensed repository. Setting the PORT variable serves it over streamable HTTP instead, for clients that need that.