Find Daily Free Tools on Product Hunt

Purpose

Surface 5–10 products from Product Hunt's current daily launches that are worth sharing with people who rely on free tools — i.e. products whose pricing label is Free or Free Options (freemium / has a usable free tier). For each product the skill returns name, tagline, topics, vote count, the canonical product URL, and the pricing label that justifies the "free" classification. Read-only — never signs in, votes, comments, follows, or submits anything.

When to Use

• A daily "free tools roundup" feed: pull the day's top launches and keep only the free / freemium ones.
• Curating a newsletter or social post of free tools for a budget-conscious audience.
• Any flow that needs today's Product Hunt launches filtered by price (the leaderboard itself does not expose pricing).
• Substitute daily with weekly / monthly / yearly leaderboards (same parsing) for longer-horizon roundups.

Workflow

Product Hunt sits behind a Cloudflare managed challenge, and there is no usable public data path without an OAuth token (the official GraphQL API at api.producthunt.com/v2/api/graphql requires a developer access token; the plain HTTP/fetch path — even /robots.txt — returns the Cloudflare "Just a moment…" interstitial). The honest, reliable method is a stealth browser session. A bare session is blocked; a stealth + residential-proxy browserless_agent call clears the challenge consistently. There is one unavoidable extra cost: the daily leaderboard and homepage do not show pricing per card, so the "free" decision requires opening each candidate's /products/{slug} page and reading its pricing label.

1. Use a stealth + residential-proxy session (mandatory): call browserless_agent with a top-level proxy: { "proxy": "residential", "proxyCountry": "us" }. Batching the whole leaderboard → per-product flow inside one call's commands array is the convenient default — it saves round-trips and avoids accidentally dropping the session config. But the session persists across calls, keyed by proxy: repeat the same proxy arg on any follow-up call to reconnect to the same warmed session (Cloudflare cookie intact), while dropping or changing proxy puts you back on the blocked bare path. There's no explicit release step to manage.

2. Open today's daily leaderboard. Leaderboards run on US/Pacific time; build the dated URL accordingly:

   https://www.producthunt.com/leaderboard/daily/{yyyy}/{m}/{d}

   (e.g. .../daily/2026/6/20). The homepage https://www.producthunt.com/ "Top Products Launching Today" section is an equivalent fallback. Then:

   { "method": "goto", "params": { "url": "https://www.producthunt.com/leaderboard/daily/2026/6/20", "waitUntil": "load", "timeout": 45000 } }
   { "method": "waitForTimeout", "params": { "time": 3000 } }

   (the waitForTimeout covers Cloudflare + progressive render). Confirm you're through Cloudflare: the page title should be like Best of Product Hunt: June 20, 2026 | Product Hunt, not Just a moment....

3. Extract the ranked launch list with a text/html read of body (or fold the parse into an evaluate command). Each launch row renders as:

   [1\. WorkClaw](/products/workclaw)Collaborative, proactive AI coworkers who work in Slack
   [Productivity](/topics/productivity)•[Artificial Intelligence](/topics/...)•[Business](/topics/...)
   40
   208

   Parse per row: rank (the N\. prefix), name, slug (from /products/{slug}), tagline (text immediately after the product link), topics (the /topics/ links), and the two trailing integers — the larger is the vote count, the smaller is the comment count. Skip Promoted rows that have no rank number (sponsor slots).

4. Classify pricing per product (2 turns each). The label is not on the leaderboard — open each candidate top-down (highest votes first), all within the same browserless_agent call:

   { "method": "goto", "params": { "url": "https://www.producthunt.com/products/{slug}", "waitUntil": "load", "timeout": 45000 } }
   { "method": "text", "params": { "selector": "body" } }

   In the returned text, the pricing/availability label sits immediately before the string Launch tags:. It is one of:

   • Free → free: true
   • Free Options (freemium / free tier) → free: true
   • Paid / Payment Required → free: false
   • (no label before Launch tags:) → unknown; skip and move on.

5. Stop early. As soon as you have 5–6 confirmed free / Free Options products, stop opening pages and emit the JSON. Six is a complete result; collecting more just burns turns/cost.

6. Emit the JSON in the Expected Output shape. There's no explicit session-release step — the session persists across calls keyed by proxy, so nothing needs to be torn down.

Site-Specific Gotchas

• Stealth + residential proxy is mandatory. A bare goto and the plain fetch/curl path both return the Cloudflare Just a moment... managed challenge (the homepage returns HTTP 403 on direct fetch; even /robots.txt is challenged). A browserless_agent call with proxy: { "proxy": "residential" } clears it reliably. Always verify the page title before trusting any extracted content.
• No free API path. The official Product Hunt GraphQL API (api.producthunt.com/v2/api/graphql) needs an OAuth developer token — out of scope for an unauthenticated agent. Don't waste time trying the API or fetch; the browser is the only reliable surface.
• Pricing is NOT on the leaderboard/homepage cards. You must open /products/{slug} and read the label before Launch tags:. This is the single biggest cost driver — budget ~2 turns per product and stop at 5–6 free hits.
• /llms.txt is open and documents every leaderboard URL pattern. https://www.producthunt.com/llms.txt (loads fine in the stealth session) gives the canonical paths: daily /leaderboard/daily/{yyyy}/{m}/{d}, weekly /leaderboard/weekly/{yyyy}/{ww} (ISO week), monthly /leaderboard/monthly/{yyyy}/{mm}, yearly /leaderboard/yearly/{yyyy}, products /products/{slug}, categories /categories/{slug}. It also states the attribution requirement: "Data sourced from Product Hunt (https://www.producthunt.com/)".
• Leaderboards are Pacific-time. Build the dated daily URL from the current US/Pacific date, or you'll fetch the wrong (or an empty/future) day. The homepage "Top Products Launching Today" tracks the same day if you'd rather not compute the date.
• Promoted sponsor rows are interleaved into the list without a rank number — they are ads; skip them.
• An Interactive tag can sit next to the pricing label on the product page (e.g. ...InteractiveFree OptionsLaunch tags:...). Anchor on the text immediately before Launch tags: so you don't mistake Interactive (a demo flag) for a price.
• Some products show no pricing label (e.g. open-source dev tools like "Mellum by JetBrains" surfaced only Launch tags: with no preceding price). Treat as unknown and skip rather than guessing free.
• Don't use snapshot for the list — a text/html read of body (or an evaluate that parses the rendered rows) renders the ranked rows far more cleanly and cheaply than walking the accessibility tree.
• Vote vs. comment count: the two trailing integers per row are comment-count then vote-count; the larger one is votes. Votes are the better "worth sharing" signal.

Expected Output

Validated 2026-06-20 (two consecutive runs converged on identical classifications). Vote counts are a snapshot — they climb during the day.

{
  "success": true,
  "source": "https://www.producthunt.com/leaderboard/daily/2026/6/20",
  "date": "2026-06-20",
  "count": 6,
  "products": [
    {
      "rank": 1,
      "name": "WorkClaw",
      "tagline": "Collaborative, proactive AI coworkers who work in Slack",
      "url": "https://www.producthunt.com/products/workclaw",
      "topics": ["Productivity", "Artificial Intelligence", "Business"],
      "votes": 208,
      "pricing": "Free Options",
      "free": true
    },
    {
      "rank": 2,
      "name": "Reframe",
      "tagline": "Surf like it's 1999",
      "url": "https://www.producthunt.com/products/reframe-7",
      "topics": ["Open Source", "User Experience", "GitHub"],
      "votes": 149,
      "pricing": "Free",
      "free": true
    },
    {
      "rank": 3,
      "name": "Slackbot's MCP Client",
      "tagline": "Work across 20+ apps in Slack with multiplayer collaboration",
      "url": "https://www.producthunt.com/products/slack",
      "topics": ["Slack", "Task Management", "Artificial Intelligence"],
      "votes": 144,
      "pricing": "Free Options",
      "free": true
    },
    {
      "rank": 5,
      "name": "pumaDB",
      "tagline": "a small hosted memory layer for AI agents",
      "url": "https://www.producthunt.com/products/pumadb",
      "topics": ["Developer Tools", "Artificial Intelligence", "Database"],
      "votes": 120,
      "pricing": "Free",
      "free": true
    },
    {
      "rank": 6,
      "name": "Foyer",
      "tagline": "Build a room of ambient sound that lives in your notch",
      "url": "https://www.producthunt.com/products/foyer-4",
      "topics": ["Mac", "Productivity", "Music"],
      "votes": 101,
      "pricing": "Free Options",
      "free": true
    },
    {
      "rank": 7,
      "name": "Are you in the Weights?",
      "tagline": "Find out if you live forever in the brain of the LLMs",
      "url": "https://www.producthunt.com/products/are-you-in-the-weights",
      "topics": ["Artificial Intelligence", "Tech", "Games"],
      "votes": 100,
      "pricing": "Free",
      "free": true
    }
  ],
  "error_reasoning": null
}

Failure shape (Cloudflare wall not cleared, or fewer than 5 free products found that day):

{
  "success": false,
  "source": "https://www.producthunt.com/leaderboard/daily/2026/6/20",
  "date": "2026-06-20",
  "count": 0,
  "products": [],
  "error_reasoning": "Stuck on Cloudflare 'Just a moment...' — browserless_agent call was missing the residential proxy."
}