Signal filtering
Before anything is stored, each incoming email passes an intelligence filter. Only messages that look like real sales signals move forward. Noise is dropped so your workspace stays focused on deal-relevant context.What typically passes the filter
| Signal type | Example | Captured? |
|---|---|---|
| Interested or engaged replies | Prospect asks a question or responds to your outreach | Yes |
| Thread replies (even brief) | “Let’s talk next week” in reply to your email | Yes |
| Real meeting signals from people | Request, acceptance, or decline from a contact (not automated system mail) | Yes |
| Objections and pushback | Concerns that affect whether the deal moves | Yes |
| Customer lifecycle mail | Renewals, expansions, complaints, or new needs from existing customers | Yes |
| New stakeholders | Introductions or forwards to someone who can decide | Yes |
| Business documents | Attachments like PDFs, Word docs, spreadsheets, decks (metadata is noted; see below) | Yes |
| Useful out-of-office | OOO with a return date or a redirect to another person | Yes |
| Warm intros | Introductions from mutual contacts | Yes — high priority |
What is usually discarded
| Noise type | Example |
|---|---|
| Marketing and bulk mail | Newsletters, promos, digests from vendors |
| Tool automation | CRM alerts, generic scheduling confirmations |
| Billing noise | Receipts and invoices unrelated to the sales relationship |
| Internal-only chatter | Team email (unless forwarded for external deal context) |
| Inbound spam / cold to you | Unsolicited cold outreach to you |
| Social notifications | LinkedIn and similar social notification emails |
| FYI CC | You’re only CC’d with nothing directed at you |
| Meaningless auto-replies | “Thanks, we got your message” with no substance |
| System calendar mail | Machine-generated accepted / declined / tentative responses |
Classification and extraction run in the language you configured during onboarding — summaries and labels read naturally in that language.
What Klipy remembers from each email
For messages that pass the filter, Klipy stores structured sales intelligence, not the full email text. Think of it as a concise memory of the conversation:| What Klipy remembers | What it means for you |
|---|---|
| Summary | A short, business-focused paragraph (up to about 100 words) focused on impact, stakes, and what should happen next — written as if a teammate briefed you. |
| Conversation type | How the message fits the sales motion — for example cold outreach, follow-up, proposal, meeting request or recap, objection handling, introduction, renewal discussion, support, or general correspondence. |
| Sentiment | The emotional tone: positive, neutral, cautious, negative, or urgent. |
| Main intent | One plain sentence on what the sender is trying to accomplish (for example, “Replying to the proposal with pricing concerns”). |
| Open concerns | Objections or blockers that could stop the deal from advancing if unaddressed. |
| Questions you still need to answer | Specific questions or requests from the sender that expect a reply or decision. |
| Buying signals | Signals like explicit interest, timelines, budget hints, authority, or willingness to move forward. |
Meetings and calls
For meetings, Klipy uses the notetaker to join calls you connect, capture audio for processing, and produce structured outputs: transcript with speaker labels, summary, action items, framework alignment, and an optional follow-up draft. Long-term retention of raw recordings is not the product model — the value is in the extracted summary and follow-on artifacts, not an archive of full recordings. For setup and behavior details, see Meeting notetaker.Messaging channels
For LinkedIn, WhatsApp, and Telegram, messages are captured as interaction records and surface in Unibox in real time when new activity arrives. When you first connect a channel, Klipy pulls recent history so you are not starting from zero: LinkedIn about the last 30 days, WhatsApp about the last 14 days. Telegram messages are also stored as interaction records in Unibox with the same real-time surfacing model.What Klipy does not capture
- Full email bodies — only the extracted intelligence described above.
- Attachment contents — filenames, types, and presence may be noted; Klipy does not ingest and store the full text of files inside your attachments.
- Private or blocked calendar events — nothing outside what you allow.
- Channels you didn’t connect — no data from sources you have not linked.
- Routine internal team mail — unless it is forwarded or clearly tied to external deal context.
Related guides
Connecting channels
Link email, calendar, meetings, and messaging so Klipy can process the right conversations.
Follow-ups & drafts
How extracted intelligence turns into draft replies you review and send.