Recordings
| Tool | What it does |
|---|---|
conversation_keyword_match | Find which saved contacts, or which external company email domains, were on the recorded calls where certain phrases were said - or not said - across the recordings the user can see. |
recording_ask | Ask a question about what was said in one recording’s meeting, such as ‘what did they say about pricing’ or ‘did they mention a competitor’, answered from its transcript. |
recording_create_followup_tasks | Create reviewed tasks from a recording’s follow-up suggestions. |
recording_draft_followup | Create a follow-up email prefill from one recording; this sends nothing. |
recording_get | Open one recording: who hosted and attended, when and how long, its AI outcome, next steps, objections, questions and decisions, talking-point hits, notes and who has access. |
recording_search | Search or filter the team’s recordings library by title, guest or participant name and email, company, host, status, tag, talking point, contact, date range, or a phrase said in the recording. |
recording_transcript | Read one recording’s transcript line by line - who said what and when - a page at a time, continuing from the returned next_offset. |
conversation_keyword_match
Find which saved contacts, or which external company email domains, were on the recorded calls where certain phrases were said - or not said - across the recordings the user can see.
Which people or companies said X across our calls: resolve an Include-ANY / Include-ALL / Exclude phrase filter over the team’s recording transcripts, narrowed to recordings visible to the caller, into the saved contacts who attended them (SavedPerson oids) or their external attendee email domains.
| Endpoint | POST https://api.prospeo.io/api/v2/tools/conversation-keyword-match |
| MCP tool | conversation_keyword_match on https://mcp.prospeo.io |
| Effect | Reads data. |
| Credits | Free, uses no credits. |
| Permission | No role permission beyond API access. |
| Confirmation | Not needed. |
| Pagination | Returns one page. |
Request
Send POST with your API key in the X-KEY header and a JSON body. Unknown fields and query parameters are rejected. The full JSON Schema is served with the tool list.
| Parameter | Type | Required | Description |
|---|---|---|---|
match | string (enum) | Yes | people returns the saved contacts who attended the matching recordings; companies returns the external email domains of their attendees. One of: people, companies. |
include_any | array of string | No | Match recordings whose transcript contains ANY of these phrases (each up to 100 characters), up to 25 phrases. Up to 25 items. |
include_all | array of string | No | Match recordings whose transcript contains ALL of these phrases (each up to 100 characters), up to 25 phrases. Up to 25 items. |
exclude | array of string | No | Exclude recordings whose transcript contains any of these phrases (each up to 100 characters), up to 25 phrases. Up to 25 items. |
Response
{"ok": true, "summary": "...", "payload": {...}} on success. Errors return {"ok": false, "code", "message", "retryable"}; see error codes.
recording_ask
Ask a question about what was said in one recording’s meeting, such as ‘what did they say about pricing’ or ‘did they mention a competitor’, answered from its transcript.
Ask a natural-language question about one recording’s meeting, answered ONLY from its diarized transcript, with client-held (never persisted) conversation history. Returns answered:false with a locked refusal line when the answer isn’t grounded in the meeting, so a genuine miss, a guard hit, and a rate-limit read identically.
| Endpoint | POST https://api.prospeo.io/api/v2/tools/recording-ask |
| MCP tool | recording_ask on https://mcp.prospeo.io |
| Effect | Reads data. |
| Credits | Free, uses no credits. |
| Permission | No role permission beyond API access. |
| Confirmation | Not needed. |
| Pagination | Returns one page. |
Request
Send POST with your API key in the X-KEY header and a JSON body. Unknown fields and query parameters are rejected. The full JSON Schema is served with the tool list.
| Parameter | Type | Required | Description |
|---|---|---|---|
recording_id | string | Yes | The recording’s oid, from a prior recording_search or recording_get result. Pattern ^[0-9a-fA-F]{24}$. |
question | string | Yes | The question to answer from this recording’s transcript, up to 500 characters. Length 1 to 500. |
history | array of object | No | Prior turns of this same conversation, oldest first, up to 12; never persisted server-side. Up to 12 items. |
Response
{"ok": true, "summary": "...", "payload": {...}} on success. Errors return {"ok": false, "code", "message", "retryable"}; see error codes.
recording_create_followup_tasks
Create reviewed tasks from a recording’s follow-up suggestions.
Create the reviewed ‘suggested tasks’ rows (drawn from a recording’s AI next steps, or custom rows) as real Task objects (Source=AI_RECORDING), each linked back to the recording. Idempotent per next-step: a row already promoted to a Task on a prior submit is silently skipped, not duplicated.
| Endpoint | POST https://api.prospeo.io/api/v2/tools/recording-create-followup-tasks |
| MCP tool | recording_create_followup_tasks on https://mcp.prospeo.io |
| Effect | Changes data. |
| Credits | Free, uses no credits. |
| Permission | No role permission beyond API access. |
| Confirmation | Not needed. |
| Pagination | Returns one page. |
Request
Send POST with your API key in the X-KEY header and a JSON body. Unknown fields and query parameters are rejected. The full JSON Schema is served with the tool list.
| Parameter | Type | Required | Description |
|---|---|---|---|
recording_id | string | Yes | MongoDB ObjectId Pattern ^[0-9a-fA-F]{24}$. |
tasks | array of object | Yes | Up to 50 items. |
Response
{"ok": true, "summary": "...", "payload": {...}} on success. Errors return {"ok": false, "code", "message", "retryable"}; see error codes.
recording_draft_followup
Create a follow-up email prefill from one recording; this sends nothing.
Draft an AI-composed follow-up email for a recording: From the host’s connected mailbox, To the primary external participant, Subject ‘Recap and next steps - {meeting name}’, Body an AI recap that never includes a recording link. Returns a composer PREFILL only — creates and sends NOTHING; the human presses Send in the compose window.
| Endpoint | POST https://api.prospeo.io/api/v2/tools/recording-draft-followup |
| MCP tool | recording_draft_followup on https://mcp.prospeo.io |
| Effect | Changes data. |
| Credits | Free, uses no credits. |
| Permission | No role permission beyond API access. |
| Confirmation | Not needed. |
| Pagination | Returns one page. |
Request
Send POST with your API key in the X-KEY header and a JSON body. Unknown fields and query parameters are rejected. The full JSON Schema is served with the tool list.
| Parameter | Type | Required | Description |
|---|---|---|---|
recording_id | string | Yes | MongoDB ObjectId Pattern ^[0-9a-fA-F]{24}$. |
Response
{"ok": true, "summary": "...", "payload": {...}} on success. Errors return {"ok": false, "code", "message", "retryable"}; see error codes.
recording_get
Open one recording: who hosted and attended, when and how long, its AI outcome, next steps, objections, questions and decisions, talking-point hits, notes and who has access.
Load one recording’s single-recording view: the list-row facts (title, status, host, times, duration, participants, artifact flags) plus the detail-only blocks — attendee roster with CRM contact links, host notes, AI insights (outcome, next steps, objections, pain points, questions, topics, decisions), talking-point hits and the People-with-access list.
| Endpoint | POST https://api.prospeo.io/api/v2/tools/recording-get |
| MCP tool | recording_get on https://mcp.prospeo.io |
| Effect | Reads data. |
| Credits | Free, uses no credits. |
| Permission | No role permission beyond API access. |
| Confirmation | Not needed. |
| Pagination | Returns one page. |
Request
Send POST with your API key in the X-KEY header and a JSON body. Unknown fields and query parameters are rejected. The full JSON Schema is served with the tool list.
| Parameter | Type | Required | Description |
|---|---|---|---|
recording_id | string | Yes | The recording’s oid, from a prior recording_search result. Pattern ^[0-9a-fA-F]{24}$. |
Response
{"ok": true, "summary": "...", "payload": {...}} on success. Errors return {"ok": false, "code", "message", "retryable"}; see error codes.
recording_search
Search or filter the team’s recordings library by title, guest or participant name and email, company, host, status, tag, talking point, contact, date range, or a phrase said in the recording.
Search and filter the team’s recordings library: merged free-text search (meeting titles, guest/participant names+emails, company names, and spoken words — with a highlighted transcript snippet on a speech hit), plus host/status/source/tag/talking-point/participant/date-range/keyword-phrase-group facets, all AND-composed and visibility-scoped. Use whenever the user names any criterion; use recording_list for a bare browse.
| Endpoint | POST https://api.prospeo.io/api/v2/tools/recording-search |
| MCP tool | recording_search on https://mcp.prospeo.io |
| Effect | Reads data. |
| Credits | Free, uses no credits. |
| Permission | No role permission beyond API access. |
| Confirmation | Not needed. |
| Pagination | Page-based: page. |
Request
Send POST with your API key in the X-KEY header and a JSON body. Unknown fields and query parameters are rejected. The full JSON Schema is served with the tool list.
| Parameter | Type | Required | Description |
|---|---|---|---|
page | integer | No | Page number, starting at 1, 25 recordings per page. From 1 to 100000. Default 1. |
search | string | No | Free text over meeting titles, guest/participant names and emails, company names, and spoken words, up to 200 characters. Length 0 to 200. |
recording_oids | array of string | No | Resolve a known set of recording oids in one page, bypassing pagination, up to 200. Up to 200 items. |
host_ids | array of string | No | Restrict to recordings hosted by these teammates, by user oid, up to 100. Up to 100 items. |
statuses | array of string (enum) | No | Stored recording-status values to include (the full 11-value set, not the collapsed display tag); empty means no status filter. Up to 11 items. Items one of: SCHEDULED, JOINING, IN_LOBBY, RECORDING, PROCESSING, ANALYZING, COMPLETED, SKIPPED, NOT_ADMITTED, CANCELED, FAILED. |
sources | array of string (enum) | No | Recording origin to include: PROSPEO or GONG; empty means no source filter. Up to 2 items. Items one of: PROSPEO, GONG. |
tag_ids | array of string | No | Restrict to recordings carrying any of these tag oids, up to 100. Up to 100 items. |
talking_point_ids | array of string | No | Restrict to recordings that hit any of these talking-point oids, up to 100. Up to 100 items. |
contact_ids | array of string | No | Restrict to recordings a saved contact participated in, by SavedPerson oid, up to 100. Up to 100 items. |
date_from | string | No | Only include recordings at or after this UTC timestamp (ISO 8601; the recording’s effective date). |
date_to | string | No | Only include recordings at or before this UTC timestamp (ISO 8601). |
keyword_include_any | array of string | No | Match a recording whose transcript contains ANY of these phrases (each up to 100 characters), up to 25 phrases. Up to 25 items. |
keyword_include_all | array of string | No | Match a recording whose transcript contains ALL of these phrases (each up to 100 characters), up to 25 phrases. Up to 25 items. |
keyword_exclude | array of string | No | Exclude recordings whose transcript contains any of these phrases (each up to 100 characters), up to 25 phrases. Up to 25 items. |
sort_direction | string (enum) | No | Sort direction ASC or DESC over the recording’s effective date; omitted keeps newest-first. One of: ASC, DESC. |
include_facets | boolean | No | Whether to also compute and return facet counts for this search; adds cost, so defaults to false for a tool call. Default false. |
Response
{"ok": true, "summary": "...", "payload": {...}} on success. Errors return {"ok": false, "code", "message", "retryable"}; see error codes.
recording_transcript
Read one recording’s transcript line by line - who said what and when - a page at a time, continuing from the returned next_offset.
Read one recording’s diarized transcript — structured lines with start offset, speaker and text, never bytes — a bounded page at a time from a line offset, with next_offset to continue. Empty when no READY transcript exists yet.
| Endpoint | POST https://api.prospeo.io/api/v2/tools/recording-transcript |
| MCP tool | recording_transcript on https://mcp.prospeo.io |
| Effect | Reads data. |
| Credits | Free, uses no credits. |
| Permission | No role permission beyond API access. |
| Confirmation | Not needed. |
| Pagination | Offset-based: offset. |
Request
Send POST with your API key in the X-KEY header and a JSON body. Unknown fields and query parameters are rejected. The full JSON Schema is served with the tool list.
| Parameter | Type | Required | Description |
|---|---|---|---|
recording_id | string | Yes | The recording’s oid, from a prior recording_search or recording_get result. Pattern ^[0-9a-fA-F]{24}$. |
offset | integer | No | Zero-based index of the first transcript line to return; pass the previous result’s next_offset to continue reading. From 0 to 100000. Default 0. |
Response
{"ok": true, "summary": "...", "payload": {...}} on success. Errors return {"ok": false, "code", "message", "retryable"}; see error codes.