The agent retrieves structured quotes, speakers, dates, and timestamped links from YouTube and podcast transcripts relevant to a specific query, enabling deep-dive analysis of media sentiment and claims.
The agent first calls arcmira_describe to load the client reference. It then constructs and executes a JavaScript program using arcmira_execute_read to perform a search with specific parameters (query, time window, limit). The agent parses the returned passages, ensuring source links and attribution are retained, and synthesizes a sourced answer or structured data for further analysis. It checks for truncation and adjusts limits if necessary.
- Category: Research and knowledge
- Effect: Read-only
- MCP servers: Arcmira: YouTube Transcript Search (https://koskamo.com/explore/arcmira-mcp)
- Source: Agents - Arcmira API (https://arcmira.com/docs/mcp-server)
- Source quote: "The agent calls arcmira_describe once to load the client reference, then runs a program with arcmira_execute_read ."
- Playbook page: https://koskamo.com/explore/use-cases/search-transcripts-for-investment-research
Prompt: Search YouTube and podcast transcripts for 'GPU financing' over the last 60 days. Return the relevant quotes with speakers when available, publication dates, transcript sources, and timestamped links. Read the surrounding exchanges before summarizing what the sources imply.