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.
72 tools total. All read-only unless otherwise noted. All scoped to the team bound to your token. Discovered automatically via tools/list — this page is for human reference.
Catalog & discovery (call these first)
These tools return the names of events, properties, agents, and spans that exist for your team. Always call them before passing names to analytical tools — guessing names returns empty results.
| Tool | Scope | Returns |
|---|
list_event_names | mcp:events | Distinct event_name values tracked by the team. |
list_event_properties | mcp:events | Property names available on a given event. |
list_property_values | mcp:events | Distinct values for one event × property (confirm exact spellings before filtering). |
list_agent_names | mcp:agent_runs | Distinct agent_name values in agent runs. |
list_agent_run_dimensions | mcp:agent_runs | Tool names, LLM model names, and retrieval span names used inside runs. |
Events analytics — mcp:events
| Tool | Returns |
|---|
run_insights_query | Time-series counts, uniques, sums, averages, or other aggregations for a single event over a chosen window, with optional breakdown and filters. |
run_funnel_query | Multi-step conversion funnel for an ordered event sequence: per-step user counts and conversion rates. |
run_retention_query | Cohort retention matrix for a starting + return event over day/week/month intervals. |
run_flow_query | Sankey flow diagram around an anchor event: the N events before and M events after it, showing where users come from and go next. Returns the query DSL for use in create_report flow chart cards. |
get_segment_comparison | Two user segments compared across one or more event metrics. |
get_utm_attribution_analysis | Event aggregated by UTM source, medium, campaign, term, content. |
get_top_users | Users ranked by event count, distinct sessions, or active days. |
get_session_analysis | Session totals, unique users, p50/p95 duration, bounce rate, sessions-per-user histogram. |
get_property_distribution | Top values + cardinality + null % for a property across events or users. |
compare_periods | Same metric across two non-overlapping windows in one call. |
get_event_correlation | P(B given A) within a session or per user, plus baseline + lift. |
get_anomaly_detection | Daily z-scored series for a metric; flags points exceeding a threshold. |
get_cohort_compare | Two predicate-defined cohorts compared on a chosen KPI. |
get_path_analysis | Most common events before / after / surrounding a target event in the same session. |
UX & technical health (also mcp:events)
| Tool | Returns |
|---|
get_rage_click_analysis | Rage-click counts per page or element, affected user counts, top targets. |
get_scroll_depth_analysis | Scroll depth per page or page-element, reach %, average max depth. |
get_exit_intent_analysis | Exit-intent rates per page over a window. |
get_form_abandonment_analysis | Form completion rate, drop-off rate, drop-off field per page. |
get_js_error_analysis | JavaScript exception counts grouped by class + message, affected users. |
get_error_impact_on_funnel | Per-step error rates joined with funnel definition, impact on conversion. |
get_page_performance_analysis | Core Web Vitals (LCP, CLS, INP) + load-time percentiles per page. |
get_network_error_analysis | Network/HTTP error counts grouped by status, host, path. |
Agent runs & traces — mcp:agent_runs
| Tool | Returns |
|---|
list_agent_runs | Paginated list of agent runs, filterable by agent, status, time, signals, cluster. |
get_agent_run | One run + full span tree (LLM/tool/agent spans), feedback, conversation siblings. |
search_agent_runs | Runs whose recorded inputs/outputs semantically match a free-text query. |
get_failed_user_attempts | Runs that semantically match a query AND show failure signals (errored, high rage, negative feedback). |
get_run_metrics | Total runs, error rate, p50/p95 duration, total cost, total tokens. |
get_token_cost_breakdown | Token usage + cost grouped by agent, model, provider, or tool. |
get_tool_call_analysis | Per-tool span metrics: call count, error rate, p50/p95 duration, avg cost. |
get_top_failure_modes | Most frequent error classes from agent spans. |
get_agent_feedback_summary | Positive/negative counts, avg rating, sample of recent comments. |
Clusters — mcp:cluster
| Tool | Returns |
|---|
list_use_case_clusters | Use-case clusters with label, run count, error rate, scores, analysis JSON. |
get_cluster_runs | Member runs of one cluster, optionally grouped by conversation. |
get_cluster_summary | Cluster row + top tools used by members + top error types. |
Issues (v2) — mcp:cluster
Typed multi-source issue detection. Replaces the deprecated list_issue_clusters.
| Tool | Returns |
|---|
list_issues | All detected issues for the team: type, severity, title, affected runs/users. |
get_issue_details | Full details for one issue including root-cause analysis. |
get_issue_timeline | Time-series of when an issue first appeared, worsened, or resolved. |
get_issue_members | Paginated list of run IDs, span IDs, or signal IDs linked to a specific issue. Use after get_issue_details to page through the full affected set. |
get_top_failing_tools | Tools with the highest error rates and their associated issue clusters. |
get_ux_rage_hotspots | Pages/elements with elevated rage-click rates linked to detected issues. |
set_issue_status | Write — Update an issue status to acknowledged, resolved, or muted. |
Evaluations — mcp:evals
| Tool | Annotations | Returns |
|---|
list_evaluators | | All evaluators for the team: name, kind, enabled status, target kind, scoring config. |
get_evaluator | | Full config for one evaluator by name or ID. |
get_eval_results | | Paginated eval result rows: run/span id, score, passed, evaluator, timestamps. Filterable by evaluator, pass/fail, time range, and target kind. |
get_eval_results_for_run | | All eval results attached to a specific agent run. |
get_eval_results_for_span | | All eval results attached to a specific span. |
list_pending_human_evals | | Human eval queue items awaiting review, filterable by evaluator and status. |
submit_human_eval_grade | Write | Submit a pass/fail grade and optional comment for a pending human eval item. |
skip_human_eval | Write | Skip a pending human eval item without grading it (moves it out of the queue). |
create_evaluator | Write | Create a new evaluator (kinds: llm_judge, python, typescript, human, composite). |
update_evaluator | Write | Update any fields of an existing evaluator. Only provided fields are changed. Bumps version when rubric fields change. |
toggle_evaluator | Write | Enable or disable an evaluator without bumping its version. |
delete_evaluator | Write | Permanently delete an evaluator and all its results. Cannot be undone. |
test_evaluator | Write | One-shot test run against a specific run or span — bypasses sampling and filters. Use to validate a rubric before going live. |
backfill_evaluator | Write | Run an evaluator against historical runs/spans. Skips already-evaluated events. Works for all evaluator kinds; human kind enqueues grading tasks. |
get_users_with_eval_score | Requires mcp:evals + mcp:user:read_pii | Users ranked by their average eval score for a given evaluator. |
Identity (PII) — mcp:user:read_pii
| Tool | Returns |
|---|
get_user_profile | One user identity record by id/email/wallet — first/last visit, sessions, identifying fields, web3 metadata. |
get_user_journey | Chronological event timeline for one user. |
find_users | Up to N user records whose distinct id, user id, wallet, or custom properties contain a fuzzy substring. |
find_users_by_wallet | User records whose primary wallet address matches a full or partial address. Returns distinct_id and wallet metadata for use with other tools. |
get_users_for_issue | User identifiers and profile snippets for everyone affected by a specific issue. Combine with get_issue_details to investigate who is impacted. |
get_users_for_event | Users who fired a specific event within an optional time window and filter set. |
list_groups | Paginated list of group identities (companies, accounts, teams) with member counts and custom properties. |
get_group_profile | Full profile of one group: custom properties (plan, MRR, region, etc.) and list of member user identifiers. |
Reports — mcp:events
| Tool | Annotations | Returns |
|---|
create_report | Write (destructiveHint: false, creates new resource) | Creates a Trodo dashboard board and returns a shareable public URL. Sections can be text cards (markdown with tables, metrics, interpretation) or live chart cards (funnel, insights, retention, flow) — chart cards store the queryDsl and re-execute it live when the report is viewed. Mix both types in one report. |
Combined-scope
| Tool | Required scopes |
|---|
get_user_agent_runs | mcp:agent_runs + mcp:user:read_pii (joins agent runs with a user identifier) |
Automation — mcp:cluster
| Tool | Annotations | Returns |
|---|
report_heal_branch | Write — agent callback | Records the fresh branch the Heal agent pushed (status → branch_ready). The agent must never open a PR — the user opens the PR from the Trodo UI. |
tools/list — call it from your client to get JSON Schema for every parameter. The MCP Inspector (npx @modelcontextprotocol/inspector) is the easiest way to browse them interactively.
