Build a Wall of Love from Reddit, Twitter & LinkedIn Mentions

Every B2B SaaS homepage has a wall of love. Almost none of them are fresh.

The quotes were picked once, hand-curated by someone who had time, and then they sat there for a year — long after the people who said them stopped being your most relevant champions.

The problem isn't that you don't have new praise. You do. It's on Reddit threads, X posts, LinkedIn updates, Hacker News comments. The problem is finding it, picking the right ones, and getting them onto the page.

This workflow does all three.

The flow is simple:

Positive mentions → AI picks the best fit for your positioning → quotes.json on your landing page

• Pulls every positive mention of your brand from the last 12 months across Reddit, X, LinkedIn, Hacker News, YouTube, and the rest
• Uses your ICP and your homepage as the rubric — picks the quotes that best match how you actually talk about yourself
• Outputs a clean quotes.json file you can drop into any landing page, CMS, or static site

Two inputs, both you already have. Run it whenever you want a refresh.

In your MCP-compatible client (the demo uses Claude Cowork, but any client works), connect the Octolens MCP. One click.

If your brand isn't already a keyword in Octolens, the prompt will add it for you on the first run using add_keyword. Mentions take a little time to accumulate after a keyword is added, so if it's brand new you'll want to wait a few days before running the rest of the workflow.

Open the prompt below and fill in two fields. A one-line ICP description ("B2B SaaS founders at dev-tools companies who care about authentic engagement") and your homepage URL.

That's the rubric. The agent fetches your homepage and uses the language and positioning there to judge fit — so the quotes it picks sound like your site, not generic praise.

No file to maintain. If your positioning changes, your homepage already reflects it.

Paste the prompt into your client, point it at your output folder, and run it.

The agent confirms your brand is tracked, reads your homepage, fetches up to 200 positive mentions, filters out anything from your own company or employees, scores each remaining mention against your ICP and positioning, picks the top 9 with theme variety, and writes quotes.json to your folder. It also explains its picks in chat so you can sanity-check before you ship.

Review the file, push it to your site, done. Run it again whenever your wall is feeling stale — before a launch, end of quarter, whenever.

Drop this into your MCP-compatible client, replace the placeholders, and run it once.

Do the following once:

1. Use these inputs as the rubric for what makes a quote a good fit:
   - Our brand: {BRAND_NAME}
   - Our ICP: {ICP_DESCRIPTION}
     (e.g. "B2B SaaS founders and marketing leads at developer-tools
     companies — they care about authentic engagement, automation, and
     staying close to their community.")
   - Our positioning: fetch {HOMEPAGE_URL} and read it. Study how we
     describe what we do, who we're for, and the language we use.

2. Make sure the brand is tracked in Octolens:
   - Use list_keywords to check whether "{BRAND_NAME}" is already a keyword.
   - If it isn't, call add_keyword with the brand name, then stop here and
     tell the user — mentions need time to accumulate, so they should run
     this prompt again once data has built up.
   - If it is, continue.

3. Use Octolens to fetch up to 200 mentions with these filters:
   - keyword: "{BRAND_NAME}"
   - sentiment: positive
   - relevance: High
   - date range: last 12 months

4. Filter out anything from your own company or its employees:
   - Call get_workspace to get the company name, domain, and description.
   - For each mention, skip it if the author appears to be the company
     itself (e.g. an official brand handle, an account using the brand
     domain) or an employee.
   - When in doubt, skip. Wall-of-love quotes should clearly be from
     external customers.

5. Pick the top 9 quotes. Aim for variety across themes — don't pick
   nine quotes about the same thing.

6. Write a file at {OUTPUT_FILE_PATH} containing a JSON array. Each entry:
   {
     "quote": "...",      // trimmed to 1–2 sentences, max ~280 chars
     "author": "...",     // display name
     "handle": "...",     // platform handle if available, else ""
     "platform": "...",   // x | reddit | linkedin | hackernews | youtube | etc.
     "url": "...",        // permalink to the original mention
     "date": "YYYY-MM-DD"
   }

7. Post a short summary in chat:
   - The 9 quotes picked, each with one line on why it landed
   - How many mentions were excluded as own-company or employee posts
   - Anything else notable you skipped and why

8. Leave the file in the folder for human review. Do not commit or push —
   the user will read the summary, sanity-check quotes.json, and ship it
   themselves.

If fewer than 9 strong matches exist, write what you found, note the count,
and stop. Don't pad with weak quotes.

• Auto-Create Linear Issues from Reddit, Twitter & LinkedIn Mentions — the first post in this MCP series.
• Octolens MCP — every tool you can drive Octolens with in plain language.