Your AI agent,
on any website.

A Thread is a small file that teaches your AI exactly how one website works — its buttons, forms, quirks, and flows. Install it, tell your agent what to do, walk away.

Affiliate signups, book publishing, quest farming, freelance gigs, email campaigns — pre-mapped by people who already use those sites daily. Your agent stops guessing and just runs it.

66 sites live · $4–12 one-time · free updates forever · engine free

11 founder-proven55 open bountiesfree updates forever

Some things APIs will never automate.

APIs cover the compute side. WebLoom covers the human-shaped half that doesn't fit into any API.

  • 01The 14 fields you have to fill in to claim an affiliate link
  • 02Posting to Twitter past the free API's 2-tweets-a-day cap, on your own account
  • 03Submitting a bid on Upwork that wants 200 chars of “why you” and a custom rate
  • 04Re-uploading a product image on Gumroad because the UI keeps eating it
  • 05Switching a setting buried 4 menus deep in a SaaS admin you visit once a quarter
  • 06Filling out the same profile across LinkedIn, GitHub, Substack, Dev.to, Bsky

The list grows every day. Every gap becomes a Thread. Every Thread cuts 30 to 90 minutes off someone's day.

01Buyer
Install a Thread.

Pick a website you use, buy its Thread once, your AI agent gets fast on it. $4–12 one-time. Yours forever. Engine is always free.

02Author
Make a Thread.

Use a site you already know daily, let WebLoom record your moves, publish the Thread, earn 75% of every sale — for as long as you keep it green.

№ 02 — how it works

A plugin that
learns as you use it.

WebLoom is a free plugin for Claude Code, Cursor, Claude Desktop, and any AI client that supports MCP. It adds 62 browser-control tools and a learning loop that records every successful action you take.

01
Drop the install file into your AI chat

Download webloom-install.md. Drag it into Claude Code, Cursor, Continue, Cline, or any MCP-aware client. Your AI reads the file, clones the engine, makes the folders, edits the MCP config. 62 browser tools live on the next message.

Download webloom-install.md
⬇ webloom-install.md
       ↓
  drag into AI chat
       ↓
  ✓ engine cloned
  ✓ ~/.webloom/ folders created
  ✓ MCP config updated
  ✓ 62 tools live
02
Use websites — normally

Tell your AI to do whatever you'd normally do — sign up for an affiliate program, publish a book, claim a quest, fill a profile. Every successful click + fill + upload silently records to your playbook file. No extra step. Your daily work IS the data.

> Sign me up for HubSpot's affiliate program.

   …agent fills the form…

playbook file updates automatically:
  ✓ Fill business info     [cdp]  ×3
  ✓ Verify email           [js]   ×1
  ✓ Submit application     [cdp]  ×1
  ✓ Configure payout       [cdp]  ×2
03
Tell your AI: package this as a Thread

When you're done driving the site, one line in your AI chat bundles every successful action into a single Thread file. Selectors, click strategies, file uploads, field types, wait times — all in one .thread.json file you can share or sell.

> Package my work on hubspot.com as a Thread.

  ✓ 7 proven actions bundled
  ✓ saved to ~/.webloom/threads/
        hubspot.com.thread.json

   ← this file IS the Thread
04
Drag the file to the Atelier — earn 75%

Drag your hubspot.com.thread.json onto the Atelier upload box. Set a one-time price ($5–12 typical based on complexity). Listed instantly. Buyers pay once, you earn 75% per sale — forever, for every buyer.

~/.webloom/threads/
  └── hubspot.com.thread.json
                ↓
         drag onto Atelier
                ↓
  ✓ live at /thread/hubspot.com
  ✓ 41 installs · $287 last month
№ 04 — install
One file.
One drop.
Engine live.

Download webloom-install.md and drag it into your AI chat. Claude Code, Cursor, Continue, Cline — any MCP-aware client reads the file and runs the install for you. No copy-paste, no manual config.

After install — where everything lives
~/.webloom/
  ├── engine/             ← the MCP server
  ├── threads/            ← your Threads land here
  │     └── *.thread.json
  ├── playbook.json       ← your learning
  └── logs/
Drop this into your AI chatrecommended

webloom-install.md

1.2 KB · drag into Claude / Cursor / Continue

Download install file
Your AI installs everything. 62 tools live on next message.docs →
or: copy the install prompt instead
Install WebLoom MCP for me.

1. Clone the engine:
   git clone https://github.com/webloomhq/engine ~/.webloom/engine

2. Install deps:
   cd ~/.webloom/engine && pip install -r requirements.txt

3. Create playbook dirs:
   mkdir -p ~/.webloom/threads ~/.webloom/logs

4. Add this to my MCP config (~/.mcp.json under "mcpServers"):
   "webloom": {
     "type": "stdio",
     "command": "python",
     "args": ["~/.webloom/engine/server.py"]
   }

After that, 62 browser-control tools become available and a local playbook
at ~/.webloom/ fills automatically as I drive websites. Any Threads I drop
in ~/.webloom/threads/ merge into the playbook at runtime.