GitHub

All notable changes to Metrx are documented in this file.

[Unreleased]

Planned Features (Phase 2)

Production gateway deployment to Cloudflare
Stripe billing activation
Advanced analytics with ML-powered insights
Team collaboration and role-based access
Budget forecasting
SDK auto-instrumentation for popular frameworks
Custom report builder
API rate limiting dashboard

[0.2.0] - 2026-02-22

MVP Enhancement Release

Added — Budget System

Budget CRUD with daily/monthly periods and microcents precision
3 enforcement modes: alert_only, soft_block, hard_block
Gateway-level budget enforcement via KV
Budget spending cache with period tracking
Budget reset cron worker
Settings UI at /settings/budgets

Added — Alert Infrastructure

Complete alert system with 5 alert types (cost_spike, error_rate, usage_spike, idle_agent, budget_warning)
Alert preferences per organization (email, Slack, Telegram, webhook)
Alert cron worker for periodic threshold checks
Email templates for alert notifications
Spending anomaly detection worker
Settings UI at /settings/alerts

Added — Dashboard Preview Mode

All dashboard pages render with synthetic data for unauthenticated visitors
Preview banner component indicating sample data
API routes return preview data when auth fails
Enables “Try the Dashboard” marketing CTA

Added — IDE-Assisted Onboarding

Modal dialog for generating AI IDE setup prompts
Supports Claude Code, Cursor, Codex, Windsurf, Gemini, Grok
Context-aware prompt generation (template, agent name, gateway URL)
Copy-to-clipboard with Python/TypeScript/JavaScript support

Added — Harness Intelligence

5 AI agent anti-pattern detectors
Context overflow, doom loop, model mismatch, no-verification, time-budget waste
Session timeline visualization
Agent detail flyout with insights

Added — Homepage Enhancements

Live topology hero visualization (animated SVG network)
Before/After comparison section
Interactive demo CTA
“Why Now?” urgency section with market stats
Enhanced social proof bar with logo placeholders

Added — Supabase Realtime

Realtime subscriptions on key tables
Custom React hooks for live updates
No-polling dashboard updates

Added — Design Partner Program

Partner registration and feedback APIs
Partner feedback widget
Design partner simulator for testing
Comprehensive test harness (200-agent seed, smoke tests)

Added — Database Schema Expansions

13 total migrations (up from 3)
Tables for: harness insights, design partners, alerts, budgets, worker health, notification channels

Fixed

Topology page now accessible via sidebar navigation
Cancel Subscription button connected with two-click confirmation
“YC Startups” removed from homepage social proof
Budget/alerts/analytics API routes added to middleware public allowlist
Preview data fallbacks for all API endpoints

Infrastructure

Worker health monitoring system
Dead letter queue for failed background jobs
Queue health checks
Multi-channel notification delivery

[0.1.0] - 2026-02-20

Initial Release (MVP)

This is the first public release of Metrx — “The P&L for Your AI Workforce.”

Added

Gateway (Cloudflare Worker)

LLM API proxy supporting OpenAI, Anthropic, Google, and xAI
Transparent request forwarding (drop-in replacement for direct API calls)
API key authentication with KV caching for sub-millisecond auth
Request/response streaming support
Automatic token counting and cost calculation
Rate limiting per organization tier
CORS support with configurable allowed origins
Custom headers for tracking: X-Agent-ID, X-Session-ID, X-Customer-ID
Request size validation (max 10MB)
Provider timeout handling (60s limit)
Health check endpoint

Web Dashboard (Next.js)

Real-time cost tracking and dashboards
Agent management and performance metrics
Session drill-down and analysis
Outcome tracking and business result management
Team management (invite users, manage roles)
Billing integration with Stripe
Webhook configuration and management
Usage vs. plan limit visualization
API key generation and rotation
Organization settings
Dark mode support

Background Workers (BullMQ)

Asynchronous event processing from Redis queue
Database writes with Supabase
Monthly usage counter updates
Session aggregation and analytics
Webhook event delivery with retry logic
Error handling and logging

Database (PostgreSQL)

Organizations table with tier management
API keys with SHA-256 hashing
Events table with full LLM call details
Sessions table for grouping related calls
Outcomes table for business result tracking
Monthly usage counters for rate limiting
Webhooks table for event subscriptions
Row-Level Security (RLS) for data isolation
Comprehensive indexes for fast queries

Authentication & Authorization

Clerk OAuth integration for dashboard
API key-based authentication for Gateway
Organization-scoped access control
Row-Level Security in database

Billing Integration

Stripe subscription management
Monthly call limits per tier
Usage-based overage detection
Webhook handling for Stripe events
Customer portal integration

SDK (@metrxbot/sdk)

Node.js client library for Gateway integration
TypeScript support
OpenAI SDK compatibility
Async/await API
Streaming support
Error handling

Supported LLM Models

OpenAI: GPT-4, GPT-4 Turbo, GPT-3.5 Turbo
Anthropic: Claude 3 (Opus, Sonnet, Haiku)
Google: Gemini Pro
xAI: Grok
OpenAI Embeddings: text-embedding-3-small, text-embedding-3-large

Pricing Tiers

Starter: 1K calls/month (Free)
Lite: 10K calls/month ($29/mo)
Pro: 100K calls/month ($99/mo)
Business: 1M calls/month ($499/mo)
Enterprise: Unlimited (Custom pricing)

API Endpoints

Gateway:

GET /health — Health check
GET /v1/models — List supported models
POST /v1/chat/completions — OpenAI-compatible chat
POST /v1/messages — Anthropic-compatible messages
POST /v1/embeddings — Embeddings endpoint

Web API:

GET /api/health — Health check with service status
GET /api/agents — List agents
GET /api/agents/{id} — Get agent details
GET /api/agents/{id}/metrics — Agent metrics
GET /api/agents/{id}/sessions — Agent sessions
POST /api/webhooks/generic/{orgId} — Register webhook
GET /api/outcomes — List outcomes
POST /api/outcomes — Create outcome
GET /api/dashboard — Dashboard data

Webhook Events

event.created — Sent when LLM call completes
outcome.confirmed — Sent when outcome is confirmed
rate_limit_approaching — Sent at 80% of monthly quota
cost_spike_detected — Sent when cost exceeds threshold

Documentation

API Reference with examples
Integration Guide with quickstart
Architecture documentation
Self-Hosting Guide
This Changelog

Security Features

TLS 1.3 for all communications
SHA-256 hashing for API keys
CORS restrictions (configurable)
Row-Level Security in Supabase
HMAC-SHA256 webhook signatures
Rate limiting and request validation
Request size limits

Observability

Detailed logging in Gateway
Sentry error tracking integration
OpenTelemetry support (spans exported to observability backends)
Health check endpoints
Request latency tracking
Error rate monitoring

Known Limitations

Maximum request size: 10MB
Maximum gateway response latency: <100ms p95
Streaming responses don’t include cost until completion
Only supports providers with Bearer token authentication
Webhook delivery is best-effort (no persistence between restarts)

Fixed

N/A (Initial release)

Security

See SECURITY_REVIEW.md for detailed security analysis
Key findings addressed in this release

Release Versioning

Metrx uses Semantic Versioning (MAJOR.MINOR.PATCH):

MAJOR: Breaking changes or major features
MINOR: New features or improvements
PATCH: Bug fixes

Future Versions

0.2.0 (Q2 2026)

0.3.0 (Q3 2026)

ML-powered prompt optimization
Budget forecasting
Competitive pricing analysis
Team collaboration features
Advanced RBAC (Role-Based Access Control)

1.0.0 (Q4 2026)

Getting Help

Documentation: docs.metrxbot.com
Issues: GitHub Issues
Email: support@metrxbot.com
Slack: Community Slack

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

License

Metrx is available under the MIT License. See LICENSE file for details.

Last Updated: 2026-02-20

Self-Hosting