BYO keys (BYOK), by design
You bring your own API keys, so there is no central account a platform can cut off and no usage we meter or mark up.
AI-filtered signal · Built for developer relations
BackTalk watches Hacker News, Reddit, GitHub, Stack Overflow, Bluesky, YouTube and more. AI filters the noise. You get one inbox, plus a daily or weekly briefing that runs your week.
7-day free trial · no credit card required
Best way to handle your-product webhooks in Next.js? Docs feel scattered…
Shipped payments with your-product this weekend, smoother than expected.
Your digest is ready: 47 mentions, 3 need attention, sentiment +12.
Built for developer advocates and DevRel teams, founders doing their own outreach, and maintainers who want to hear their community without watching a dozen tabs. And nothing covert: BackTalk never posts as you or joins communities you don’t own. Replies happen on the native platform, by you, via deep link.
For developer relations
Developers rarely @-mention you; they just talk. BackTalk listens where they actually are, and turns the firehose into a handful of moments that need a human reply. Catch these conversations as they happen, or batch them into a daily or weekly briefing. Here is what that looks like on a normal week.
Someone hits a wall wiring up your SDK on Stack Overflow or Ask HN. BackTalk surfaces it with a relevance score and a deep link, so you can reply as yourself, on the platform, before the thread goes cold.
A developer vents about a broken webhook on Reddit. It lands in your needs-attention queue, tagged and prioritised, so you can route it to engineering and close the loop in public.
A team is weighing you against an alternative on Hacker News, or asking "what should I use for X?" You see it as it happens and join the conversation while your answer can still change the outcome.
A tutorial, a launch thread, a kind word on dev.to or YouTube. Star it, thank the author, and grow the advocate network that quietly compounds into your best channel.
One digest, daily or weekly, your call: mentions found, sentiment trend, response rate, and what is still unanswered. Exec-ready numbers that make the case for your team when budgets are set.
Point Claude at your own BackTalk.sh Inbox over MCP, ask "what did the community complain about today?", and get an answer from what BackTalk actually caught, not a fresh web search.
How it works
No dashboards to watch. Highlights what social conversations are important to you, filters, tags and groups them by what actually needs your reply.
Add keywords once. BackTalk.sh polls every source with cursors and dedupe, so nothing is fetched twice and nothing slips through.
Every match runs through AI relevance and intent checks. Job ads and crypto spam die here. Questions, complaints, praise and leads get tagged.
One inbox with needs-attention on top. A daily or weekly digest that writes your status report for you and your team.
The AI
BackTalk's AI does the first pass then hands you the short list that actually needs you.
The wrong "Prisma" (the band, not your database), job ads and crypto spam are gone before they reach you. Every mention that survives comes with a relevance score, so you only read what's actually about you.
Each mention is scored positive, neutral or negative, and your digest plots the trend with the threads that moved it. So a bad week is a number you can see instead of a feeling.
An AI rank from urgent to low puts the churn-risk complaint at the top and the nice-to-have at the bottom. You answer what needs a reply in hours, not whatever happened to be newest.
Every conversation is labelled (question, bug report, complaint, praise or comparison) and rated for intent from high to low. Filter to just the bug reports for engineering, or just the high-intent posts where a deal is live.
Run the filtering on our AI, or plug in your own LLM key (Claude, for instance). You get the newest models on your own keys, and we never use your content to train anyone's model.
An MCP server lets Claude or any agent query your own BackTalk inbox. Ask "what did developers complain about today?" and the answer comes from what BackTalk caught, not a fresh web search.
The product
Dashboards demo well and retain badly. BackTalk.sh is built around the two surfaces you will actually open.


Coverage
Including the ones nobody can monitor. If a platform is closed, we say so instead of selling you a checkmark.
Hacker News
posts and comments
your own API key, or via Brave
GitHub
issues, discussions, code
dev.to
posts and comment threads
Substack
via Brave web index
Medium
via Brave web index
public posts only, via Brave web index
Stack Overflow
questions and answers across the Stack Exchange network
YouTube
video titles and descriptions, official API
Bluesky
public posts via the official AppView search, no key needed
RSS / any blog
Mastodon
Lobsters
Web discovery
any blog or docs site
Product Hunt
Discord (your server)
Slack (your community)
npm / PyPI stats
X / Twitter
their prices, not ours
Compare
Every tool below is good at what it was built for. This table shows where each one listens and how each one is built, so you can pick the right tool, even if it is not ours.
| BackTalk.shAI dev-signal monitoring | Octolensreal-time DevRel monitoring | Common Roomenterprise community intelligence | Brand24broad social listening | Syftenindie keyword alerts | F5Botfree email alerts | |
|---|---|---|---|---|---|---|
| Where it listens | ||||||
| Hacker News | Yes | Yes | No | No | Yes | Yes |
| Yes | Yes | Yes | Yes | Yes | Yes | |
| GitHub | Yes | Yes | Yes | No | Yes | No |
| Stack Overflow | Yes | Yes | Yes | No | Stack Exchange | No |
| dev.to | Yes | Yes | Yes | No | Yes | No |
| public posts | Yes | your own page only | limited | No | No | |
| Substack & Medium | Yes | newsletters | No | as blogs | as blogs | No |
| YouTube | Yes | Yes | chosen channels only | Yes | Yes | No |
| X / Twitter | No | Yes | No | Yes | No | No |
| Bluesky | Yes | Yes | No | No | No | No |
| How it is built | ||||||
| Bring your own API keys | Yes | No | No | No | No | No |
| MCP server for AI agents | Yes | Yes | No | No | No | No |
| Pricing | flat, from $19/mo | metered, from $159/mo | enterprise, from $2,100/mo | monthly subscription | monthly subscription | free |
Compiled from each vendor's public documentation. Spotted an error? Email us and we will fix it.
Built different
You bring your own API keys, so there is no central account a platform can cut off and no usage we meter or mark up.
Seen, replied, ignored, starred. Cursors, dedupe and thread-follow underneath. State is the product, and it is exactly what a scheduled AI prompt can never give you.
BackTalk.sh never posts as you, never joins communities you do not own, never resells data. Replies happen on the native platform, by you, via deep link.
An MCP server on every plan. Ask Claude what the community complained about today or this week and get an answer from your own BackTalk.sh Inbox, not a fresh web search.
Pricing
Unlimited mentions with no overage bills: you bring your own keys, so there's no usage.
For a developer keeping an ear on their own project.
$19/month
For the DevRel-of-one covering products, SDKs and competitors.
$99/month
For a DevRel team sharing one inbox.
$250/month
Every plan starts with a 7-day free trial. No credit card required. Cancel anytime.
FAQ
Including the awkward ones.
Public posts. LinkedIn has no official API for keyword monitoring. BackTalk.sh covers what can be covered cleanly: public LinkedIn posts surface through the Brave Search web index, fetched with your own Brave API key, and land in your inbox next to everything else.
A lot of the work. Every mention runs through the AI relevance filter. What survives is scored for sentiment (positive / neutral / negative), ranked by priority from urgent down to low so you know what needs a reply within hours, and categorised (question / bug report / complaint / praise / comparison). You can run it on our AI or bring your own LLM key, Claude for instance, so you get the newest models on your own terms.
A scheduled AI prompt is stateless: every run re-searches from zero, has no idea what you already saw or replied to, and rides web search indexes that lag by hours or days. Monitoring is a state problem. BackTalk.sh keeps per-source cursors, deduplicates everything, tracks replied and ignored, and spends AI tokens only on genuinely new items. Your agent is still welcome: the MCP server lets Claude or ChatGPT query your BackTalk.sh Inbox directly.
Each platform (Reddit, Brave Search, your LLM) is accessed with your own developer key instead of one giant central account. That means no surprise bills, no usage we can meter or mark up, and if one source ever objects, only that source pauses.
No. Every reply happens on the native platform itself via a deep link by you. BackTalk.sh never auto-posts, never joins communities you do not own, and never monitors private spaces.
No. We never resell content or per-person data, both because platform terms forbid it and because it would be a terrible thing to build a trust product on. We also never use your workspace content to train AI models.
Spin up your inbox in minutes. AI filters the noise, scores what matters, and hands you the handful of mentions that actually need you.
7-day free trial · no credit card required.