Analytics Dashboard

A personal developer performance dashboard — not a monitoring tool. Track productivity, cost, and reliability across all AI coding providers. All data stays on your machine.

Open it: Ctrl+Shift+P -> MR Multi coder: Open Analytics Dashboard

Dashboard Preview

Analytics Dashboard — filters, productivity score ring, and 12 KPI cards with period-over-period deltas

Filters, productivity score ring with component bars, and 12 KPI cards with period-over-period deltas

Usage Trends and Productivity Trend charts

Usage Trends & Productivity Trend

Peak Productivity Hours and Efficiency Over Time charts

Peak Hours & Efficiency Over Time

Session Depth, Session Quality, Activity Heatmap, and Task Categories

Session Depth, Quality, Heatmap & Task Categories

Daily Cost Trend, Cost Efficiency, and Cost by Project

Cost Intelligence — Daily Trend, Efficiency & By Project

Cost by Provider, Model Ranking, Error Rate, and Context Compaction

Model Ranking, Error Rate & Context Compaction

Response Time by Provider, Model Usage, and Token Breakdown

Response Time, Model Usage & Token Breakdown

How Data Is Collected

All analytics data comes from a local database stored on your machine. No data ever leaves your machine.

TEXT

Provider turn completes
  -> AnalyticsCollector records event
    -> AnalyticsStore inserts row into SQLite
      -> Dashboard queries the same DB on open

What Gets Recorded

Field	Source
`timestamp`	Wall clock at response completion
`session_id`	Current chat session
`project_id`	Workspace folder name
`provider`	Active provider (claude, kilo, qwen, etc.)
`model`	Model identifier (e.g. claude-sonnet-4-6)
`prompt_tokens`	Tokens in the request
`response_tokens`	Tokens in the response
`response_time_ms`	End-to-end latency
`status`	success or failure
`task_category`	debugging, refactoring, code_generation, documentation
`accepted`	Whether the suggestion was accepted

Privacy

- No source code is stored. Only metadata (token counts, timing, status).
- No API keys or credentials are ever logged.
- Fully opt-out: set mrMultiCoder.analytics.enabled to false.
- Local-only storage. The database is never uploaded, synced, or shared.

Productivity Score

A single composite metric (0-100) answering "How productive am I with AI assistance?" Displayed as a colour-coded ring with five component bars.

Component	Weight	What It Measures
Consistency	20%	How regularly you use the tool
Acceptance	25%	How often you keep AI suggestions
Reliability	25%	How error-free your interactions are
Efficiency	15%	How fast the AI responds
Depth	15%	How engaged your sessions are

Colour thresholds: Green (80+), Blue (60-79), Yellow (40-59), Red (<40).

KPI Cards

Twelve key metrics displayed as cards. Each card shows the current value and a delta arrow with percentage change vs. the equivalent previous period.

KPI	Notes
Total Prompts	Total AI interactions
Total Sessions	Distinct conversations
Current Streak	Consecutive active days ending today
Longest Streak	Best streak in the date range
Success Rate	% of interactions without errors
Acceptance Rate	% of suggestions the user kept
Time Saved	Estimated hours saved (configurable minutes per task)
Total Cost	Estimated USD spend — decrease is green
Cost/Accepted	Cost per accepted suggestion
Projected /mo	Daily average x 30
Active Days	Days used out of range
Avg Response	Mean response latency — decrease is green

Productivity Charts

Usage Trends

Line chart showing prompts and sessions per day. Answers: "Am I using the tool more or less over time?"

Productivity Trend

Dual-axis chart: prompt volume (bars) overlaid with success rate % (line).

Peak Productivity Hours

24-hour chart showing requests, accepted suggestions, and acceptance rate by hour.

Efficiency Over Time

Tokens per accepted suggestion vs. acceptance rate — downward trend means better prompting.

Session Depth Distribution

Bar chart bucketing sessions by turn count: 1, 2-5, 6-10, 11-20, 21-50, 50+.

Session Quality

Four cards classifying sessions: Excellent (70%+ acceptance, <10% errors, 3+ turns), Good, Fair, Poor.

Activity Heatmap

7x24 grid (day of week x hour) with opacity-scaled cells.

Task Categories

Pie chart: debugging, refactoring, code generation, documentation, uncategorized.

Cost Intelligence

Daily Cost Trend

Area chart with gradient fill showing estimated daily spend.

Cost Efficiency

Summary card: Total Cost, Total Prompts, Accepted, Cost/Prompt, Cost/Accepted.

Cost by Project

Horizontal bar chart ranking projects by spend (top 8).

Cost by Provider

Vertical bar chart comparing cost across providers.

Cost Estimation Rates

Model	Rate (per 1K tokens)
Claude Sonnet 4.6	$0.003
Claude Opus 4.8	$0.015
Claude Haiku 5.5	$0.0001
Qwen3 Coder Plus	$0.0004
Default (all others)	$0.002

Free-tier providers (Kilo, OpenRouter free models, Ollama) show zero cost.

Performance & Reliability

Model Productivity Ranking

Sortable table ranking every model by composite score (60% acceptance + 40% speed).

Error Rate Trend

Dual-axis chart: error count (bars) and error rate % (line).

Context Compaction

Tracks total compactions, sessions with compaction, and daily frequency.

Response Time by Provider

Horizontal bar chart comparing average response time across providers.

Model Usage

Pie chart showing request distribution across models.

Token Breakdown

Six-metric summary: avg/max prompt tokens, avg response tokens, total tokens, retry rate.

Goal Setting & Tracking

Set personal targets in VS Code settings under mrMultiCoder.analytics.goals.*

Setting	Default	What It Tracks
`acceptanceRate`	0 (off)	Target acceptance rate — shown as a goal line
`dailyPrompts`	0 (off)	Target daily prompts — shown as a goal line
`maxCostPerDay`	0 (off)	Cost ceiling — shown as a limit line
`maxResponseTimeMs`	0 (off)	Response time target — shown as a limit line

When active, a Goal Status Bar appears with colour-coded chips and KPI cards get coloured borders. Set any goal to 0 to disable.

Filters & Export

Global Filters

Filter	Options	Default
Date Range	Today, Last 7/30/90 Days, Last 6 Months, This Year, Custom	Last 7 Days
Project	All Projects, or any workspace in the data	All
Provider	All Providers, or any provider with events	All
Model	All Models, or any model (filtered by provider)	All

Export

Two formats available via header buttons: CSV (flat table of all raw events) and JSON (structured format). Both respect current filters.

Additional Features

Period-over-Period Comparison

Every trend chart includes a dashed overlay showing the equivalent previous period for visual comparison.

Real-Time Auto-Refresh

Dashboard refreshes every 30 seconds by default (configurable). A green pulsing dot indicates live status.

Timeline Annotations

Mark important events (e.g. "Deployed v2.0") on timeline charts. Stored in SQLite, persists across sessions.

Calibrated Time Saved

Default: 5 minutes per accepted suggestion. Configurable via mrMultiCoder.analytics.minutesPerTask (1-60 minutes).

Drill-Down

Click any data point on Usage Trends to see every session for that date with full details.

Analytics Settings

Setting	Default	Description
`mrMultiCoder.analytics.enabled`	true	Enable/disable analytics data collection
`mrMultiCoder.analytics.minutesPerTask`	5	Minutes saved per accepted suggestion (1-60)
`mrMultiCoder.analytics.refreshIntervalSeconds`	30	Auto-refresh interval (0 = disabled, max 300)
`mrMultiCoder.analytics.goals.acceptanceRate`	0	Target acceptance rate (0-100%). 0 = off
`mrMultiCoder.analytics.goals.dailyPrompts`	0	Target daily prompts. 0 = off
`mrMultiCoder.analytics.goals.maxCostPerDay`	0	Max daily cost in USD. 0 = off
`mrMultiCoder.analytics.goals.maxResponseTimeMs`	0	Max response time in ms. 0 = off