Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.trodo.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

Trodo MCP uses five scopes. Four are on by default; the PII scope is opt-in.

The five scopes

ScopeDefaultDescription
mcp:events✓ ONQuery event counts, funnels, retention, sessions, breakdowns, anomalies, cohort comparisons, paths. No personally-identifying data.
mcp:agent_runs✓ ONList + inspect AI-agent runs, their span trees, error rates, costs, feedback.
mcp:cluster✓ ONBrowse use-case + issue clusters of agent conversations.
mcp:evals✓ ONList evaluators and results, manage the human eval queue, submit grades, and create new evaluators.
mcp:user:read_pii✗ OFFLook up individual users by identifier or fuzzy match. Returns emails, wallet addresses, country, device. Off by default.

Tool → scope mapping

mcp:events

run_insights_query, run_funnel_query, run_retention_query, get_segment_comparison, get_utm_attribution_analysis, get_top_users, get_session_analysis, get_property_distribution, compare_periods, get_event_correlation, get_anomaly_detection, get_cohort_compare, get_path_analysis, get_rage_click_analysis, get_scroll_depth_analysis, get_exit_intent_analysis, get_form_abandonment_analysis, get_js_error_analysis, get_error_impact_on_funnel, get_page_performance_analysis, get_network_error_analysis, list_event_names, list_event_properties, list_property_values, create_report

mcp:agent_runs

list_agent_runs, get_agent_run, search_agent_runs, get_failed_user_attempts, get_run_metrics, get_token_cost_breakdown, get_tool_call_analysis, get_top_failure_modes, get_agent_feedback_summary, list_agent_names, list_agent_run_dimensions

mcp:cluster

list_use_case_clusters, get_cluster_runs, get_cluster_summary, list_issues, get_issue_details, get_issue_timeline, get_top_failing_tools, get_ux_rage_hotspots, report_heal_branch

mcp:evals

list_evaluators, get_evaluator, get_eval_results, get_eval_results_for_run, get_eval_results_for_span, list_pending_human_evals, submit_human_eval_grade, create_evaluator

mcp:user:read_pii

get_user_profile, get_user_journey, find_users, find_users_by_wallet, list_groups, get_group_profile

Combined scopes

  • get_user_agent_runs — requires both mcp:agent_runs AND mcp:user:read_pii (it joins agent runs with a user identifier).
  • get_users_with_eval_score — requires both mcp:evals AND mcp:user:read_pii (returns user identities ranked by eval score).

Choosing scopes

For most analytics use cases, the three default scopes are enough. Claude can answer questions about engagement, conversion, agent performance, and cluster trends without ever seeing user PII. Enable mcp:user:read_pii if you specifically want Claude to investigate individual users (e.g. “what did user [email protected] ask my agents this week”). The consent screen calls this out separately and the API key UI requires checking the box explicitly. mcp:evals is on by default because viewing eval results and managing the human eval queue are core analytics workflows with no PII exposure (evals are keyed by run/span IDs, not user identities).

Scope changes

For OAuth tokens: re-running the consent flow with different scope checkboxes issues a fresh token family with the new scopes. The old grant remains revocable from Integrations → MCP for Claude & Cursor. For direct API keys: scopes are fixed at creation. Revoke and re-issue to change them.