Create a Meeting Prep & Follow‑Up Bot with Amazon Quick & Webex MCP

Problem

Professionals spend too much time digging through past meeting transcripts, email threads, and video highlights to get ready for an upcoming call. After the call, they often scramble to capture decisions and action items. The result is missed context, duplicated effort, and slower execution.

What if a single prompt could pull the agenda, relevant past discussions, and unanswered follow‑ups, then hand you a concise brief before the meeting, and later produce a clean summary with action items?

According to the AWS Machine Learning Blog, Amazon Quick can be combined with Cisco Webex MCP servers to create exactly that kind of assistant.

Prerequisites

• Amazon Quick workspace – an account with permissions to create agents and access external data sources.
• Cisco Webex MCP server – the organization’s Webex Meeting Content Platform (MCP) must be reachable via API keys.
• API credentials – a Webex integration token that allows read‑only access to meetings, transcripts, Vidcast highlights, and message threads.
• Basic scripting ability – Python or Node.js to glue API calls together; the guide uses Python examples.
• IAM role – a role that lets Amazon Quick invoke external APIs securely.

Steps

1. Set up the data connectors

In the Amazon Quick console, add a new Webex MCP connector. Provide the Webex integration token and specify the endpoint URLs for meetings, transcripts, Vidcast, and message threads. Quick will treat each endpoint as a searchable data source.

Test the connection by issuing a simple query like list upcoming meetings for user@example.com. The response should be a JSON list of scheduled meetings.

2. Define the assistant’s prompt template

Create a prompt that tells Quick what to do when a user asks for a prep brief. The template can look like this:

"You are a meeting assistant. When given a meeting ID, retrieve the meeting time, agenda, any prior summaries, transcript excerpts, and Vidcast highlights. Also scan Webex message threads for unresolved follow‑ups. Compile a brief that includes:
- Meeting title and time
- Key topics from past discussions
- Relevant video clips
- Open items from chat"

Save the template as PrepBriefPrompt.

3. Build the pre‑meeting workflow

Write a small function that accepts a calendar event or a meeting ID. The function will:

1. Call the Webex MCP /meetings endpoint to confirm the meeting exists.
2. Pull the latest meeting summary and transcript using the /transcripts endpoint.
3. Request Vidcast highlights via /vidcast/highlights.
4. Search the /messages endpoint for threads that reference the meeting ID and filter for messages marked "unresolved".
5. Feed all collected snippets into the PrepBriefPrompt and let Amazon Quick generate the brief.

Return the generated brief to the user via Slack, Teams, or email.

4. Deploy the function as a Quick agent

In the Quick console, create a new agent called WebexPrepAgent. Attach the IAM role created earlier, link the PrepBriefPrompt, and point the agent’s execution handler to the function from step 3. Enable the trigger onUserPrompt so the agent activates whenever a user types something like /prep meeting 12345.

5. Test the pre‑meeting assistant

Pick a meeting scheduled for tomorrow. Run the command /prep meeting 98765. Verify that the output includes:

• Meeting title, date, and time.
• Bullet points summarizing the last discussion.
• Links to the most relevant Vidcast clips.
• A short list of pending chat items.

If anything is missing, adjust the data‑source queries or prompt wording and re‑run.

6. Create the post‑meeting workflow

The after‑meeting assistant uses a similar pattern but a different prompt. Define PostSummaryPrompt such as:

"You are a meeting recorder. Summarize the discussion from the provided transcript, highlight decisions, and list action items with owners. Keep the summary under 300 words."

When a meeting ends, a Webex webhook can fire a POST request to an AWS Lambda function. That function will:

1. Fetch the final transcript and any newly uploaded Vidcast clips.
2. Pass them to the PostSummaryPrompt via Amazon Quick.
3. Send the resulting summary back to the meeting organizer’s inbox.

7. Wire the webhook to the Lambda

In the Cisco Webex developer portal, create a webhook that listens for the meeting.ended event. Set the target URL to the Lambda’s API Gateway endpoint. Ensure the Lambda has permission to invoke Quick and to read from the MCP server.

8. Automate distribution of action items

After the summary is generated, parse the list of action items (Quick can output them as a JSON array if the prompt asks). Loop through each item and post a message to the original Webex chat thread, tagging the responsible participant. This keeps the follow‑up visible where the conversation happened.

9. Monitor and refine

Enable CloudWatch metrics for the Lambda and Quick agent. Track latency (how long the brief takes to generate) and error rates (failed API calls). If latency exceeds a few seconds, consider caching the transcript or Vidcast data ahead of time.

Pro Tips

• Cache recurring data. For teams that meet weekly, store the last meeting’s transcript in an S3 bucket. Quick can read from S3 faster than pulling from Webex each time.
• Use concise prompts. The more specific the prompt, the less chance the model will hallucinate a decision that never occurred.
• Tag owners explicitly. Ask Quick to output action items in the format @username – task so the downstream chat message automatically notifies the right person.
• Limit transcript length. If a meeting runs over an hour, split the transcript into 5‑minute chunks and feed them sequentially to Quick. This avoids hitting token limits.
• Secure the webhook. Verify the signature header sent by Webex to ensure only authentic events trigger your Lambda.