Eight open sources · swept every 180 seconds

Your users already posted the bug.

People post workarounds in forum threads instead of filing bug reports. Riffle reads those threads and writes up what is actually broken.

8open sources, read for you
25findings written up a month, from $29
$0overage, at any tier

We cap what gets written up, never what gets read. A week that brings four thousand mentions still costs what your tier costs, and the report tells you it was four thousand.

Try it

Run it on any public product.

The same pipeline paying customers get, capped at three findings. No signup.

Or run withastro/astro

What a finding includes nothing running
The problem
What a user actually ran into, in one line, not a keyword match.
Their words
The sentence they wrote, quoted, so you can judge it yourself.
The link
The post it came from. One click, and you are looking at it.
Your tracker
Whether it matches an issue in your own tracker, or nothing does.

Nothing is pinned here. Run a product above and this fills in with what the pipeline actually found.

Triageclaude-haiku-4-5
Read in fullclaude-opus-5
Findings outover MCP, to your own AI
Written to your repositorynothing, ever
Hacker News · Stack Exchange · GitHub issues · Discourse forums · Lobsters · Dev.to · Mastodon · Bluesky
How it works

Every finding links to the post it came from.

Point it at your product once. After that, the reading is not your job.

01

Point it at your product

Your product name, its site or its repository, and a line on what it does. That last part is how it tells your name from a coincidence. "Riffleware" is not "Riffle", and a job ad in your category is not one of your users.

02

It reads everything, you read what matters

Every candidate gets triaged first: is this about you, and does it describe a problem someone actually hit. What survives that gets read in full, thread and replies.

03

A finding, not a notification

The problem in the user's own words, the link to the thread it came from, and the issue in your own tracker it matches if there is one. Findings come out over MCP, so the AI you already code with can read them and start on the fix.

a support forum, forty-one replies deepnever filed
anyone got this working with a symlinked content directory? mine just hangsthe opening post, which your tracker never sees
i gave up and copied the folder instead of symlinking it, works fine that waythis line is the finding
same, been doing that for monthsreply twelve, and the reason it is a blocker
What it never does

What it never does.

Four hard limits, set on purpose.

01

It never touches your repository.

We read public data and write up what we found. No commits, no pull requests, no access to your code. Your AI does the fixing, under your review.

02

It never bills above your tier.

We cap what gets written up, never what gets read. There's no overage line, at any tier, in any week.

03

It never asks you to take its word.

Every finding carries the link and the sentence the user actually wrote. If we got it wrong, one click shows you that.

04

It never touches walled-garden data.

Open sources only. Nothing scraped, nothing purchased, nothing a platform's lawyers can take away.

Pricing

Three tiers. No surprise invoice.

$29/mo Solo. One product, 25 findings written up a month. Choose Solo
Most popular $79/mo Team. One product, 90 findings, a digest every morning. Choose Team
$199/mo Studio. Three products, 270 findings, your sources read first. Choose Studio
What is in each tier ›

Every mention we find gets read. The best ones get written up, as many as your tier covers, and nothing bills above that. Pay yearly and two months are free. 30-day money-back, no questions. Cancel from the dashboard, not a support ticket. Pay, then sign in at rifflehq.in/app with the same email.