Skip to main content

​
Available Tools

Tiro MCP Server provides four tools optimized for efficient meeting data retrieval through the Progressive Disclosure pattern.

search_notes

Search meetings and return metadata only~50-100 tokens per result

get_note_summary

Get AI-generated meeting summary~200-800 tokens

get_note_transcript

Retrieve full conversation transcript~3,000-5,000 tokens

get_note_documents

Extract structured data (action items, decisions)~300-2,000 tokens

​
Tool Comparison

ToolPurposeToken UsageScope RequiredBest For
search_notesFind meetings50-100 per notemcp:notes:readInitial discovery
get_note_summaryQuick overview200-800mcp:notes:readUnderstanding key points
get_note_transcriptFull details3,000-5,000mcp:notes:readComplete information
get_note_documentsStructured data300-2,000mcp:notes:readAction items, decisions

​
Progressive Disclosure Pattern

Tiro MCP Server implements a Progressive Disclosure pattern to optimize token usage and response times.

​
The Three-Stage Approach

1

Stage 1: Discovery

Use search_notes to find relevant meetingsReturns: Note ID, title, participants, dateToken Cost: ~50-100 per noteTime: ~1 second
2

Stage 2: Summary

Use get_note_summary for key informationReturns: AI-generated summary, main topicsToken Cost: ~200-800Time: ~2 seconds
3

Stage 3: Full Details

Use get_note_transcript only when neededReturns: Complete conversation transcriptToken Cost: ~3,000-5,000Time: ~5 seconds

​
Why Progressive Disclosure?

Problem: Loading all transcripts for 10 meetings = 30,000+ tokensSolution: Load metadata first (500 tokens), then details for selected meetings onlySavings: Up to 97% token reduction
Metadata queries return in ~1 second vs ~5 seconds for full transcripts.Users get immediate results and can drill down as needed.
AI agents have limited context windows (e.g., Claude: 200K tokens).Loading only necessary data preserves context for analysis and conversation.

​
Usage Patterns

​
Pattern 1: Find and Summarize

Use Case: “What was discussed in recent marketing meetings?” Token Usage: ~1,000 tokens total (vs 15,000 with full transcripts)

​
Pattern 2: Find Specific Information

Use Case: “Find action items from last week’s sprint planning” Token Usage: ~500 tokens (vs 5,000 with full transcript)

​
Pattern 3: Deep Dive

Use Case: “Analyze entire conversation about project timeline” Token Usage: ~5,000 tokens (only when necessary)

​
Tool Details

​
search_notes

Metadata-Only SearchReturns basic information (ID, title, participants, date) without actual meeting content. Always use this first to discover meetings.
Parameters:
  • content (optional): Search term for meeting content
  • createdAt (optional): Filter by date (ISO 8601 datetime)
  • personName (optional): Filter by participant name
Returns: Array of meeting metadata Example Use: 
Find meetings about "product roadmap" from last month
Full documentation →

​
get_note_summary

Fast OverviewAI-generated summary provides key information in 80% less tokens than full transcripts. Use this before loading full transcripts.
Parameters:
  • noteId (required): Meeting note ID from search_notes
Returns: AI-generated summary text Example Use: 
What were the main points of meeting #123?
Full documentation →

​
get_note_transcript

High Token UsageFull transcripts consume 3,000-5,000 tokens for typical 1-hour meetings. Use summaries first unless you need verbatim content.
Parameters:
  • noteId (required): Meeting note ID from search_notes
Returns: Complete conversation transcript with speaker labels Example Use: 
Show me the exact conversation about budget allocation in meeting #123
Full documentation →

​
get_note_documents

Structured DataExtract organized information like action items, decisions, and key discussion points in a structured format.
Parameters:
  • noteId (required): Meeting note ID from search_notes
Returns: Structured document with sections, action items, and decisions Example Use: 
What action items came out of meeting #123?
Full documentation →

​
Best Practices

Never guess note IDs. Always use search_notes first to:
  • Discover available meetings
  • Get accurate note IDs
  • Verify meeting exists
Use get_note_summary before get_note_transcript:
  • 80% faster response time
  • 85% fewer tokens
  • Often sufficient for user needs
When looking for specific information types:
  • Action items → get_note_documents
  • Decisions → get_note_documents
  • General context → get_note_summary
  • Exact quotes → get_note_transcript
Load metadata for multiple meetings at once with search_notes, then fetch details only for selected meetings.

​
Performance Comparison

​
Traditional Approach (No Progressive Disclosure)

User: "Find last week's meetings about marketing"

Agent loads 5 meetings × 5,000 tokens = 25,000 tokens
Response time: ~25 seconds
Context used: 25,000 tokens

​
Progressive Disclosure Approach

User: "Find last week's meetings about marketing"

1. search_notes → 500 tokens, 1 second
2. User selects 2 meetings
3. get_note_summary × 2 → 1,200 tokens, 4 seconds

Total: 1,700 tokens, 5 seconds (93% reduction)

​
Next Steps

search_notes

Start here: Learn to search meetings

Authentication

Setup OAuth 2.0 access

Troubleshooting

Common issues and solutions