MCP tool optimization for conversational journalist workflowsbacklog_item

journalismmcpinvestigationux
1 min read · Edit on Pyrite

Problem

The journalist's primary agent interface is Claude Desktop/Cowork via MCP. Tool descriptions, parameter names, and output formats need to be optimized for conversational use — not just programmatic correctness.

Scope

Tool Descriptions

  • Write tool descriptions as capabilities: "Search for events involving a person or organization" not "Query investigation_event entries by actor field"
  • Include example invocations in descriptions
  • Group related tools logically for tool-use selection
  • Output Formatting

  • Structured but human-readable (not raw JSON dumps)
  • Include natural language summaries alongside structured data
  • Truncate long results with "showing 10 of 247 — use limit parameter for more"
  • Include action suggestions: "3 claims are unverified — use investigation_claims to review them"
  • Contextual Defaults

  • `kb_name` defaults to the most recently used investigation KB
  • Date ranges default to investigation's active period
  • Importance thresholds default based on investigation size
  • Session Continuity

  • MCP tools should support a "continue" pattern: "Show me more results" or "Now filter by actor X"
  • Investigation context persists across tool calls within a session
  • Acceptance Criteria

  • Tool descriptions are conversational and include examples
  • Output includes natural language summaries
  • Defaults reduce parameter verbosity for common operations
  • Journalist can work fluidly via Claude Desktop without knowing internal field names