Carbon Voice MCP

@CarbonVoice retrieve the transcript from yesterday's standup meeting and pull out action items assigned to the engineering team.

Based on the standup transcript from January 15, the engineering team has three action items: (1) Sarah will investigate the API timeout issue reported in production by end of week, (2) Marcus will review the database migration PR before tomorrow's deployment window, and (3) The team collectively agreed to schedule a technical design review for the new authentication flow next Tuesday at 2pm. The transcript shows these were confirmed verbally during the meeting's final five minutes.

This example assumes Carbon Voice has already recorded and transcribed the meeting. OAuth scope must include read access to meeting transcripts. The AI can extract structured data from conversational speech, but accuracy depends on audio quality and speaker clarity in the original recording.

@CarbonVoice start recording our current conversation and tag it as 'Q1 Planning Discussion' for the product team workspace.

Recording started for this conversation at 10:23 AM PST. The session is now being captured and will be tagged as 'Q1 Planning Discussion' in the product team workspace. Participants detected: 3 active speakers. The recording will automatically transcribe in real-time and be available for review immediately after you end the session. You can stop recording anytime by saying 'stop recording' or using the workspace controls.

This demonstrates the MCP's ability to initiate voice capture on demand. Requires OAuth write permissions and may trigger user consent flows depending on workspace settings. Be aware that recording creates persistent data — ensure all participants know they're being recorded before starting.

@CarbonVoice compare the topics discussed in our last three sprint retrospectives and identify recurring themes the team keeps mentioning.

Analyzing retrospectives from December 12, December 26, and January 9, three recurring themes emerge: (1) 'Deployment friction' mentioned in all three sessions, specifically around manual approval steps slowing Friday releases. (2) 'Documentation gaps' raised twice, with the team noting onboarding takes longer than expected for new engineers. (3) 'Cross-team dependencies' appeared in the two most recent retros, particularly regarding design handoff timing. The team's tone around deployment friction has shifted from neutral to frustrated across the three meetings.

This showcases the AI's ability to synthesize patterns across multiple voice recordings. The MCP provides transcript access; the AI handles thematic analysis. Results depend on how consistently your team labels and stores retrospective recordings. Works best when meetings follow a semi-structured format.

When voice notes replace Slack walls of text

A 6-person distributed design team uses voice memos to share context that's too nuanced for Slack threads. Carbon Voice MCP connects those recordings to the workspace so the AI can summarize them, pull action items, or answer questions about what was said last week. OAuth2 auth means each designer's recordings stay scoped to their permissions. This works when your team already records voice notes (Loom, voice messages, meeting snippets) and you need the AI to treat them like searchable documents. If your team doesn't naturally speak ideas out loud, forcing it won't help. Best fit: teams that think faster by talking than typing, and need those thoughts indexed alongside written docs.

Turn client call recordings into ticket context

A 3-person CS team records client calls and wants the AI to reference what was promised or discussed when drafting follow-up emails or updating CRM notes. Carbon Voice MCP pulls transcripts or audio metadata so the AI can cite specific moments from the call. This saves the manual step of re-listening or skimming auto-transcripts. The threshold: if your calls are under 10 minutes or already have clean written notes, the MCP adds friction. If calls run 30+ minutes and context gets lost between the recording and the ticket, this closes the gap. OAuth2 keeps each manager's call library private unless explicitly shared.

When research recordings need to inform roadmap chats

A PM and two designers conduct 8 user interviews a month. They record each session but rarely have time to re-watch before sprint planning. Carbon Voice MCP lets the AI pull quotes, themes, or pain points from those recordings when the team asks 'what did users say about onboarding?' in a workspace thread. This works if your interview volume is high enough that manual tagging falls behind, but low enough that you're not drowning in hundreds of hours of footage (where a dedicated research tool makes more sense). The buying call: if your team references past interviews more than twice a sprint, this MCP turns recordings into living context instead of archived files.