Read AI MCP

@Read AI MCP show me the key decisions and action items from yesterday's product sync meeting

Based on the transcript from your 2:00 PM product sync, three decisions were recorded: (1) Push feature freeze to next Friday to accommodate the auth refactor, (2) Allocate two engineers to the mobile performance sprint, (3) Defer the analytics dashboard redesign to Q2. Four action items were assigned: Sarah to draft the revised timeline by EOD, Marcus to audit current mobile bottlenecks by Wednesday, the design team to present Q2 roadmap options next Monday, and you to sync with stakeholders on the new feature freeze date. The meeting ran 47 minutes with 6 participants.

This example assumes Read AI has already processed and transcribed the meeting. The MCP surfaces structured data (decisions, action items, attendees) that would otherwise require manual review of a full transcript. Useful when you need meeting outcomes without re-watching recordings, but accuracy depends on Read AI's transcription quality and your team's meeting discipline in verbalizing decisions clearly.

@Read AI MCP create a summary of all engineering standups from this week and identify any blockers mentioned more than once

Across five engineering standups this week (Monday through Friday), the following blockers appeared multiple times: API rate limiting on the third-party payment provider (mentioned in 3 standups, affecting checkout flow work), staging environment instability (mentioned in 2 standups, delaying QA cycles), and unclear requirements for the user permissions redesign (mentioned in 2 standups, blocking frontend implementation). Single-mention blockers included Docker build times, a flaky test suite, and waiting on design assets. Total standup time this week: 2 hours 14 minutes across the team.

This synthesis example shows the MCP aggregating patterns across multiple meetings. It requires Read AI to have captured all referenced standups and tagged blockers consistently. The output helps surface systemic issues that might not be obvious in isolated daily check-ins. Be aware that blocker identification depends on explicit verbal mentions — unspoken or Slack-only blockers won't appear.

@Read AI MCP what topics did the leadership team spend the most time discussing in Q4 all-hands meetings?

In the four Q4 all-hands meetings, the leadership team allocated discussion time as follows: 38% on go-to-market strategy and customer acquisition (including the enterprise sales pivot announced in October), 22% on product roadmap and feature prioritization, 18% on team growth and hiring plans, 12% on operational efficiency and cost management, and 10% on company culture and remote work policies. The October all-hands had the longest discussion segment on sales strategy (24 minutes), while the December meeting focused heavily on 2024 planning. Total all-hands time in Q4: 4 hours 52 minutes.

This example demonstrates the MCP's ability to analyze meeting content over time and quantify discussion patterns. It's valuable for understanding leadership focus areas and communication priorities. However, time allocation is a proxy for importance — some critical decisions happen in brief exchanges. The analysis also assumes Read AI captured all relevant meetings and that your OAuth scope includes historical meeting data.

When meeting transcripts need to reach remote teammates fast

A 12-person startup runs weekly all-hands on Monday mornings, but three engineers are in APAC and miss the live call. The Read AI MCP pulls the transcript and action items within minutes of the meeting ending, so the remote team reads the summary before their Tuesday standup instead of waiting for someone to manually post notes. OAuth2 auth means each teammate sees only the meetings they're invited to, which matters when discussing roadmap pivots or customer churn. The trade-off: if your team already uses Notion or Confluence for meeting notes and someone owns that process, adding another tool creates duplicate sources of truth. This MCP wins when speed and automatic capture beat manual note-taking, and your team trusts AI summaries enough to skip re-watching hour-long recordings.

How sales teams close the loop on discovery calls without replaying audio

A 5-person sales team runs 8-12 discovery calls per week. The Read AI MCP lets the account exec pull the transcript and key objections into Switchy immediately after the call, then draft a follow-up email while the context is fresh. The alternative is replaying 40 minutes of audio to find the two sentences where the prospect mentioned budget timing, or relying on handwritten notes that miss exact phrasing. OAuth2 scope means the MCP only accesses calls the exec attended, so customer data stays compartmentalized. The boundary: if your team needs sentiment analysis or speaker diarization beyond basic transcription, you'll hit the MCP's limits and need the full Read AI web app. This setup works when the goal is fast follow-up, not deep conversation analytics.

When training sessions need to become searchable docs without manual editing

A 20-person agency onboards two new hires per quarter with a series of recorded Zoom trainings on client workflows, tooling, and escalation paths. The Read AI MCP pulls transcripts into Switchy where the ops lead can search across all onboarding calls to find the exact explanation of how to handle a specific client request, then paste that snippet into the internal wiki. Without the MCP, someone has to watch six hours of recordings or hope the original trainer remembers what they said three months ago. The catch: if your onboarding is already documented in Notion or Loom and people actually update it, adding meeting transcripts creates noise instead of clarity. This MCP wins when tribal knowledge lives in recordings and you need a way to surface it without assigning someone to transcribe and organize everything manually.