> ## Documentation Index
> Fetch the complete documentation index at: https://klipy.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# How do I ask about my own data?

> Ask Claude natural-language questions about your own deals and conversations — grounded in your Klipy data, with citations to the source, never the web.

You have a question that lives across a dozen conversations — what a prospect keeps hesitating on, who mentioned budget season, where a deal actually stands. The answer is in your data; finding it means re-reading everything. Connect Claude to Klipy and it answers from your own conversations, with citations back to the source.

## Example prompt

Copy this into Claude once it is connected to Klipy:

```text theme={null}
Across all my conversations with Acme, what concerns keep coming up that
I haven't fully addressed yet?
```

## What happens

<Steps>
  <Step title="Claude searches your interactions">
    Claude calls `search_interactions` to run a semantic search over your email, meeting, and social interactions, and `get_record_timeline` or `search_crm_records` to pull the relevant records — then reasons over what it finds.
  </Step>

  <Step title="Or it asks Klipy directly">
    For a synthesized answer, Claude calls `ask_klipy`, which retrieves the relevant interactions and produces one grounded answer with citations back to the source — a single read-only pass, no writes.
  </Step>

  <Step title="You get a grounded answer">
    The answer names the pattern and points to the interactions it came from, so you can verify it against the real conversations.
  </Step>
</Steps>

This is read-only — Claude reads and reasons, nothing is drafted or sent, so it consumes no tokens.

## Good to know

* **It searches your own Klipy data, not the internet.** `ask_klipy` is grounded question-answering over your organization's own interactions — email, meetings, and social threads. It does no web research and looks at no competitor's site. The answer comes entirely from conversations already in Klipy.
* **No writes, no multi-step loop.** `ask_klipy` is one retrieval-and-synthesis pass. It reads your interactions and answers — it does not take actions, edit records, or chain tool calls.
* **Answers are strongest over email and meeting content.** Semantic search is richest across your email and recorded-meeting text. Deep recall of every line of an informal chat is thinner, so lean on it most for email- and meeting-heavy questions.

## Related

<CardGroup cols={2}>
  <Card title="Pre-call research" icon="user-search" href="/docs/guides/mcp/use-cases/pre-call-research">
    Turn a question into a full pre-call account brief.
  </Card>

  <Card title="Pipeline forecast" icon="chart-line" href="/docs/guides/mcp/use-cases/pipeline-forecast">
    Ask the numbers question — a weighted view of your open deals.
  </Card>

  <Card title="Tool reference" icon="wrench" href="/docs/guides/mcp/tools">
    Every tool Claude can call, grouped by scope.
  </Card>
</CardGroup>