Screenshots
See It in Action
Dashboard — leverage trends, summary cards, and project breakdown
Features
Quantify the AI Advantage
Most teams know AI helps. Fulcrum tells you exactly how much, tracks cognitive fatigue with neuroscience rigor, and predicts where the highest leverage lies next.
Metrics & Measurement
The Numbers That Matter
AI Leverage Factor Calculation
Human estimated hours multiplied by 60, divided by Claude's actual minutes. A single number that captures the real productivity multiplier for every task — typically 20x to 120x for engineering work.
Supervisory Factor Tracking
Measures the output per minute of human supervisory investment — the time spent writing the prompt that kicked off the task. Captures the true multiplier of human-AI collaboration beyond raw task time.
Continuous Fatigue Scoring
0–1 fatigue score computed from record count (50% weight), cumulative supervisory minutes (25%), and session duration (25%). Four discrete levels: Low (0–3 records), Moderate (4–7), High (8–12), Very High (13+).
Server-Side Computed Columns
Leverage factor and supervisory factor are always recomputed server-side as generated PostgreSQL columns. No stale data, no client-side math errors, no inconsistency — the database is the single source of truth.
Analytics & Reporting
See the Full Picture
Time Series Analytics
Track leverage trends over configurable date ranges — days, weeks, months, or custom windows. Identify which types of work benefit most from AI assistance and how your productivity evolves over time.
Per-Project Breakdowns
Aggregate leverage metrics by project with record count, average leverage factor, and total hours saved. See which codebases, features, and subsystems yield the highest AI productivity gains at a glance.
Today Stats
Quick daily summary showing today's record count, average leverage, total hours saved, and current fatigue level. One glance tells you where the day stands.
React Dashboard with Recharts
Interactive dashboard with time series charts, project breakdowns, fatigue gauges, and leaderboards. Built with React 19 and Recharts for smooth, data-dense visualizations that load instantly.
Team Management with Leaderboards
Compare leverage factors across team members. Identify best practices from high-leverage engineers and share them across the organization. Friendly competition that drives better AI-human collaboration.
Export to CSV and JSON
Full data export with configurable date range and project filters. CSV for spreadsheets, JSON for programmatic analysis. Your data is always portable and never locked in.
Decision Support
Protect Your Judgment
Decision Fatigue Assessment
Based on neuroscience research into glutamate accumulation in the prefrontal cortex during sustained cognitive work. Not a vague "how tired are you" survey — a quantified model of how decision quality degrades over a workday.
Personalized Recovery Recommendations
Priority-ranked recovery suggestions with neuroscience rationale. Not generic "take a break" advice — specific interventions calibrated to your current fatigue level, time of day, and work pattern.
Agent Guidance Output
Dynamically adjusts AI agent behavior based on your fatigue level. At low fatigue, Claude asks clarifying questions. At high fatigue, it maximizes autonomy, reduces decision points, and presents completed work rather than options.
Data Management
Own Your Data
Bulk Record Import
Import historical records from CSV with automatic duplicate detection. Migrate from spreadsheets, other tracking systems, or backfill from log files without creating duplicates.
Full-Text Search
Search task descriptions across your entire history using PostgreSQL GIN indexes with tsvector. Find that task from three months ago by keyword in milliseconds, not minutes of scrolling.
Soft Deletes with Retention
Deleted records are soft-deleted with a 90-day retention window before permanent purge. Accidental deletions are recoverable. Compliance-friendly audit trail preserved.
Integration
Works Everywhere You Do
Claude Code Session Integration
Every Claude Code session automatically logs leverage records via CLAUDE.md instructions. Dual-write to both CSV and REST API ensures no record is ever lost. The most seamless AI productivity tracking possible.
CLI That Works Without a Server
The command-line interface imports the service layer directly — no running server required. Log records, query stats, and export data from a shell script or cron job on any machine with Python installed.
MCP Server — 28 Tools
Full Model Context Protocol server with 28 tools. Log leverage records, query analytics, assess fatigue, get predictions, and manage your account — all accessible from Claude Code without leaving the terminal.
API Key Authentication
Argon2id hashed API keys with prefix-based lookup for efficient authentication. Create, rotate, and revoke keys from the dashboard. Industry-standard key security without the complexity.
Rate Limiting
Configurable rate limiting at 100 requests per minute by default. Protects the API from runaway scripts or misconfigured integrations without affecting normal usage patterns.
Light and Dark Mode
Full dark mode support that follows your system preference. Charts and visualizations adapt their color palette for comfortable viewing in any lighting condition.
Prediction
Know Before You Start
Predictive Leverage Estimation
Before starting a new task, estimate its leverage factor using historical data from similar tasks. Prioritize high-leverage work first. Know where AI delivers the highest return before committing time.
Accessibility
Built for Everyone
WCAG 2.1 AA Compliance
4.5:1 contrast for body text, 3:1 for large text and UI components, in both light and dark themes.
Keyboard Navigation
Every interaction reachable via keyboard. Logical tab order, visible focus indicators, Escape-to-dismiss for modals.
Screen Reader Support
VoiceOver, NVDA, and JAWS tested. Semantic HTML, ARIA labels, live regions for dynamic updates.
Reduced Motion
Respects
prefers-reduced-motion. Usable at 200% zoom. Touch targets meet 44x44 minimum.
How It Works
From Task to Insight
Step 1: Log the Task
Record task description, human estimate, AI time, and supervisory minutes. Automatically via Claude Code or manually through the dashboard and CLI.
Step 2: Calculate Leverage
Leverage and supervisory factors computed server-side as generated PostgreSQL columns. No stale data, no client-side math errors.
Step 3: Assess Fatigue
The fatigue model evaluates cumulative cognitive load from record count, supervisory minutes, and session duration. Recovery recommendations adjust in real time.
Step 4: Optimize Workflow
Predictive estimates prioritize high-leverage tasks. Agent guidance dynamically adjusts AI behavior based on your fatigue level.
Technical Specifications
Under the Hood
Backend
- FastAPI (Python 3.12+)
- PostgreSQL with generated columns and GIN indexes
- REST API with pagination and full-text search
- Argon2id API key hashing with prefix lookup
- Soft deletes with 90-day retention
- Rate limiting (100 req/min configurable)
- MCP server (28 tools)
Frontend
- React 19
- Recharts time series and project visualizations
- Fatigue gauge and recovery recommendations
- Team leaderboards
- Light and dark mode
- Responsive design
Integrations
- Claude Code automatic session logging
- Dual-write: CSV + REST API
- CLI (works without running server)
- CSV and JSON export with filters
- Bulk import with duplicate detection
- MCP integration for AI assistants
Neuroscience
- Glutamate accumulation fatigue model
- Continuous 0–1 fatigue scoring
- Four discrete fatigue levels
- Personalized recovery recommendations
- Dynamic agent behavior adjustment
Development
100% Built by Claude
Every tool in the Renkara fleet was built by Claude (Anthropic) working alongside a single human supervisor. Every line of code, every test, every deployment: AI-authored with human direction. The leverage factor across the fleet runs in the 20x–50x range, with individual sessions regularly exceeding 100x.
See the daily leverage records for per-task numbers across the full build history.