Hi everyone — sharing a Marketplace app we built after our own shared inboxes filled up with guest-post pitches, “rank 1 on Google” offers, and other junk that still created CRM contacts.
Pain point it addresses
• Native allow/deny lists work for known senders/domains, but spam often rotates addresses with the same pitch patterns.
• Workflows can’t reliably enroll on inbound subject/body without Ops Hub custom code + Conversations API.
• Mark as spam doesn’t automatically add the sender to the deny list.
What Inbox Spam Filter does
• Connects to HubSpot Shared Inbox and scores new messages with Gemini AI (configurable threshold, default 70).
• Flags spam with score + short reasoning so agents can audit decisions.
• Optional CRM property spam_filter_status = SPAM for cleanup lists.
• Portal-wide Scoring / Notifications / Billing; allowlist + custom instructions on paid plans.
• Privacy option (Basic+): strip email addresses before AI evaluation (HubSpot records unchanged).
Who it’s for
Sales / support / CS teams on Conversations Shared Inbox drowning in outreach spam who don’t want to maintain keyword workflows or custom code.
Pricing
Starter is free (up to 500 emails/mo). Basic $25 / Pro $50 for higher volume.
Happy to answer setup questions, false-positive tuning, or how this compares to provider-level quarantine + deny lists. Screenshots of Overview + Scoring tabs coming in a follow-up reply.
Overview — spam log with score + reasoning per message
Scoring — threshold (default 70), allowlist, and custom instructions on paid plans
Typical setup path:
1. Install from Marketplace and connect the Shared Inbox(es) you want scored
2. Start with threshold 70; raise it if you see false positives on legit outreach
3. allowlist known partners/domains so they never get flagged
4. (Optional) Use spam_filter_status on contacts to build a cleanup list
Compared to native tools:
• Deny list = best for repeat senders/domains
• Provider quarantine = best to stop mail before HubSpot
• Inbox Spam Filter = helpful when pitches look the same but From addresses keep changing
Happy to walk through a false-positive example if you paste a redacted subject/snippet.
Thanks for sharing this and breaking down how it compares to native options! The framing (deny list → provider quarantine → third-party scoring) is genuinely helpful for teams trying to figure out where to start.
Thanks for sharing this. One feature that would also be valuable is giving admins a way to review false positives and use that feedback to refine future classifications.
Even with AI doing the initial filtering, having an easy review process helps teams build confidence in the results while reducing the risk of legitimate emails being missed.
HI @OCherevatyy
They make the setup and scoring workflow much easier to understand.
One feature I’d also find useful would be a review queue where admins can quickly mark messages as “Not spam” or “Spam,” with those decisions helping refine future classifications.
That would give teams more confidence in the AI while keeping the review process efficient.
That’s already the intended loop: review flagged messages on Overview (score + reasoning) or via the daily digest email, then correct future classifications via Scoring → custom instructions (plus threshold / allowlist when needed). Digest is especially handy for spotting false positives without living in the app.
Happy to help draft instruction text for a false-positive pattern if you share a redacted example.
