Analytics Dashboard

The gateway analytics dashboard provides aggregate visibility into gateway traffic, performance, and token usage over configurable time ranges. It is available as a built-in panel in the /admin web UI and through REST API endpoints under /admin/api/analytics/*.

Activation

The analytics dashboard is available on any gateway built with the admin feature (default). There is no separate configuration flag — if /admin is enabled, the analytics panel is available.

# Default: analytics included
dcc-mcp-server --app maya

# Disable admin (also disables analytics)
dcc-mcp-server --no-admin

REST API Endpoints

All endpoints require a range query parameter indicating the look-back window. Supported values: 7d, 30d (default), 90d, 180d, 365d.

GET /admin/api/analytics/overview

Aggregate KPIs for the selected time range.

{
  "range_days": 30,
  "calls_total": 15420,
  "calls_success": 14803,
  "calls_failed": 617,
  "success_rate": 0.96,
  "tokens_input": 28500000,
  "tokens_output": 4200000,
  "tokens_saved": 12300000,
  "duration_ms_avg": 842,
  "duration_ms_min": 12,
  "duration_ms_max": 45000,
  "llm_prompt_tokens": 15000000,
  "llm_completion_tokens": 3800000,
  "llm_total_tokens": 18800000,
  "unique_instances": 12,
  "unique_agents": 8,
  "daily_series": [
    { "date": "2026-05-06", "calls": 512, "failed": 18, "success": 494,
      "tokens_input": 950000, "tokens_output": 140000, "tokens_saved": 410000 },
    ...
  ],
  "top_tools": [
    { "name": "maya_scene__get_session_info", "calls": 3200, "failed": 2,
      "success": 3198, "success_rate": 0.999, "duration_ms_avg": 45 },
    ...
  ]
}

GET /admin/api/analytics/timeseries

Time-series broken down by DCC type. Supports granularity=day (default) or granularity=hour.

{
  "range_days": 30,
  "granularity": "day",
  "points": [
    { "timestamp": "2026-05-06T00:00:00Z", "calls": 512,
      "dcc_breakdown": { "maya": 320, "blender": 192 } },
    ...
  ]
}

GET /admin/api/analytics/heatmap

Hour-of-day × day-of-week heatmap data.

{
  "range_days": 30,
  "cells": [
    { "weekday": 1, "hour": 9, "calls": 142, "failed": 3 },
    { "weekday": 1, "hour": 10, "calls": 189, "failed": 1 },
    ...
  ],
  "max_calls": 234
}

GET /admin/api/analytics/export

Bulk data export. Accepts format=csv (default) or format=jsonl.

GET /admin/api/analytics/export?range=30d&format=csv
Content-Disposition: attachment; filename="analytics-export-30d.csv"

Web UI Panel

The analytics panel in /admin renders:

• Range selector: 7d / 30d / 90d / 180d / 365d
• KPI card grid: total calls, success rate, failure count, input/output/saved tokens, average duration, LLM token breakdown, unique instances, unique agents
• Daily trend mini-chart: bar chart of calls per day; failed days highlighted in red
• Heatmap: 7 (weekday) × 24 (hour) grid with blue intensity gradient; hover for tooltip with call/failure counts
• Top tools table: ranked by call volume with failure count, success rate, and average duration
• Export links: CSV and JSONL download

Data Source

Analytics are computed from AdminAuditRecord entries stored in:

1. SQLite (~/.dcc-mcp/gateway/admin/audit.db) when durable audit is configured via DCC_MCP_GATEWAY_AUDIT_DIR
2. In-memory ring buffer (default, last ~5000 records)

For persistent analytics across gateway restarts, set DCC_MCP_GATEWAY_AUDIT_DIR to a writable directory.

Data Aggregation

Aggregation runs on-demand when API endpoints are called. The core aggregate_audits() function groups audit records by (date, dcc_type, hour) into DayAggregate structs tracking:

Field	Source
calls_total	Count of audit records
calls_success	Records with success == true
calls_failed	Records with success == false
tokens_input	Sum of llm_usage.input_tokens
tokens_output	Sum of llm_usage.output_tokens
tokens_saved	Sum of context.tokens.total_saved
llm_*	Direct LLM token counters
duration_ms_*	Min/max/sum of duration_ms
instance_ids	Unique instance set
agent_ids	Unique agent set

Limitations

• Data retention is bounded by the audit store capacity (default ~5000 in-memory records; for SQLite, by DCC_MCP_GATEWAY_AUDIT_MAX_ROWS / MAX_BYTES)
• Aggregation is computed on every request; very large time ranges or high-traffic gateways may see slower response times
• The dashboard refreshes every 5 seconds via React Query polling

See Also

• admin-ui.md — full admin dashboard documentation
• gateway.md — gateway architecture and configuration
• observability.md — metrics, tracing, and Prometheus export