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

# Search sales conversations and fetch transcripts

> Search the calls recorded by Lusha Conversations for your account - metadata, AI summaries, action items, risks, and coaching analysis - then fetch speaker-attributed transcripts one conversation at a time.

The Conversations API exposes the sales calls recorded by [Lusha Conversations](/user-guide/getting-started-with-conversations/lusha-conversations-overview) to your own tooling. Search across the account's conversations, then pull the full transcript for any one of them.

It's two endpoints working together: search returns everything *about* a conversation, and a second call returns what was actually *said*.

<CardGroup cols={2}>
  <Card title="Search Conversations" icon="magnifying-glass" href="/api-reference/conversations/search-conversations">
    `POST /v3/account/conversations/search` - metadata, summaries, and analysis.
  </Card>

  <Card title="Get Conversation Transcript" icon="file-lines" href="/api-reference/conversations/get-conversation-transcript">
    `GET /v3/account/conversations/{conversationId}/transcript` - the speaker-attributed transcript.
  </Card>
</CardGroup>

## Two search modes on one contract

Search Conversations behaves differently depending on whether you supply `query`:

| Mode        | How to trigger | Behavior                                                                           |
| ----------- | -------------- | ---------------------------------------------------------------------------------- |
| **Keyword** | Supply `query` | Ranks conversations by transcript content. **All other filters are ignored.**      |
| **Filter**  | Omit `query`   | Filters structurally by dates, contact names, company domains, and meeting titles. |

Both modes return the same response shape. An empty body is valid and returns the first page of the account's conversations.

<Warning>
  In keyword mode the structural filters are ignored rather than combined. If you need both a text match and a date range, filter the results client-side.
</Warning>

## Filter-mode fields

| Field                 | Notes                                                                                       |
| --------------------- | ------------------------------------------------------------------------------------------- |
| `conversationIds`     | Return only these conversations.                                                            |
| `dateFrom` / `dateTo` | Meeting date bounds, inclusive (`YYYY-MM-DD`).                                              |
| `contactNames`        | Partial, case-insensitive match on participant names. Multiple values are ORed.             |
| `companyDomains`      | Domains of external participants (e.g. `acme.com`). Matches on domain, not display name.    |
| `meetingTitles`       | Partial, case-insensitive match on the meeting title. Multiple values are ORed.             |
| `page` / `pageSize`   | Page starts at 1. `pageSize` above 100 is rejected with `400` - it is not silently clamped. |

## What search returns

Each conversation comes back with its metadata plus AI-derived analysis: summary, action items, risks, objections, competitor mentions, coaching analysis, and chapters. **Transcripts are not included.**

<Note>
  `summary`, `coaching`, and `chapters` come from an asynchronous pipeline and are `null` or empty until it has run. A `null` here means "not processed yet", not "nothing found". Only conversations whose post-call processing has completed are returned at all.
</Note>

## Fetching a transcript

Take a `conversationId` from search and call the transcript endpoint. There's no request body and no query parameters.

<Steps>
  <Step title="Find the conversation">
    Call `POST /v3/account/conversations/search` in either mode and pick the conversation you want.
  </Step>

  <Step title="Fetch the transcript">
    Call `GET /v3/account/conversations/{conversationId}/transcript` to get the speaker-attributed, timestamped transcript.
  </Step>
</Steps>

<Warning>
  A single `404` covers all three failure cases: the conversation doesn't exist, its processing hasn't completed, or it belongs to another account.
</Warning>

<Tip>
  Transcripts can be long. If you're feeding one to an LLM, budget for the token count or summarize before passing it through.
</Tip>

## Billing

| Endpoint                    | Cost                                                                                                                                         |
| --------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------- |
| Search Conversations        | Contact your account manager for pricing on your plan.                                                                                       |
| Get Conversation Transcript | 1 credit per successful request via `ci_transcript_analysis`, including repeat requests for the same conversation. A `404` is never charged. |

While `ci_transcript_analysis` isn't seeded on your account's pricebook, the transcript endpoint stays free.

## Related pages

* [API reference: Search Conversations](/api-reference/conversations/search-conversations)
* [API reference: Get Conversation Transcript](/api-reference/conversations/get-conversation-transcript)
* [Lusha Conversations overview](/user-guide/getting-started-with-conversations/lusha-conversations-overview) - the product these recordings come from