All notable changes to Metrx are documented in this file.

[1.11.0] - 2026-03-13

Deep Audit + Tier Gating Enforcement

Added

• Tier-Based Data Access Limits — Dashboard, analytics, and agent metrics endpoints now enforce per-tier history windows (Starter: 7 days, Lite: 30 days, Pro: 90 days, Business/Enterprise: unlimited)
• Alert Tier Gating — Alert preferences and test alert endpoints now require Lite tier or higher, returning 402 with upgrade guidance for Starter organizations
• Expanded Model Coverage — Cost analysis and optimization now support 35+ models across 9 providers (OpenAI, Anthropic, Google, xAI, Meta, Mistral, Cohere, DeepSeek, Amazon)
• Enhanced Optimization Engine — 48 models tracked with 6 analyzers: model switching, waste detection, provider arbitrage, revenue intelligence, token guardrails, and predictive failure

Fixed

• Cost calculations corrected across dashboard, analytics, and optimization — all dollar amounts now display accurately
• Billing page plan limits now always match canonical tier definitions
• Python SDK duplicate event recording in OpenAI and Anthropic auto-instrumentation resolved

[1.10.0] - 2026-03-11

GTM Strategies + Gap Implementation

Added

• AI Cost Calculator (/calculator) — interactive comparison tool showing projected savings across GPT-4, Claude, Gemini, and open-source models with personalized ROI estimates
• Dynamic Model Registry — centralized model metadata (pricing, context windows, capabilities) powering the calculator and cost analysis tools
• Python SDK Auto-Detection — instrument() one-liner with automatic framework detection for LangChain, CrewAI, and AutoGen
• Cost Leak Bounty System — automated detection and notification of cost optimization opportunities
• Shareable Case Study Reports — before/after cost analysis reports with unique share URLs for social proof

Changed

• Alert email delivery improved with better template matching and error handling
• Payment failure webhook handling hardened against duplicate events

Fixed

• Alert emails not sending when Loops template IDs were mismatched
• Payment webhooks silently failing on duplicate event IDs

[1.8.0] - 2026-03-03

MCP Positioning Optimization

Added

• MCP Server expanded to 23 tools across 10 domains — query costs, run optimizations, manage budgets, configure alerts, and more — all from your AI agent or IDE
• metrx_ namespace prefix on all MCP tools for clear identification (e.g., metrx_get_cost_summary). Unprefixed aliases still work but are deprecated
• AI agent discovery files — llms.txt, llms-full.txt, agent-card.json, ucp.json so AI agents can find and understand Metrx automatically
• Smithery registry listing for one-click MCP server installation
• Rate limiting in MCP server: 60 requests/minute per tool with automatic retry and backoff
• “When NOT to use” guidance in all tool descriptions for better agent routing
• Sitemap for improved SEO and AI crawler discoverability

Fixed

• API key prefix corrected to sk_live_ across all documentation
• Customer-facing changelog synced through v1.8.0

[1.7.0] - 2026-03-02

Post-MVP UX Suite

Added

• Auto-Claim Links — secure, token-based promo claim system replacing guessable codes. Share a personalized /claim/[token] link with partners
• Setup Guide Card — 4-item onboarding checklist on dashboard (Alerts, Slack, Outcomes, Budget) with inline actions. Personalized by your technical comfort level
• Progressive Navigation — dashboard tabs unlock as you onboard. Power users see everything immediately
• Help Icons — hover over any ? icon on the dashboard for plain-English metric explanations
• Post-Onboarding Email Sequence — 3 helpful emails over your first week, each skipped if you already completed the action
• First-Login Product Tour — 5-step guided walkthrough on first dashboard visit. Replay anytime from Settings
• “Show All Features” toggle in Settings → Display

[1.6.0] - 2026-03-02

Event-Driven Architecture, Per-Agent Budgets & Unit Economics Engine

Added

• Inngest event-driven architecture — 13 background functions for attribution, optimization, alerts, anomaly detection, and more. Replaces cron-based processing for better reliability
• Per-agent budget configuration — set daily or monthly limits per agent with alert-only, soft-block, or hard-block enforcement
• Webhook field mapping — map custom CRM fields for outcome integrations
• Theme settings for dashboard customization
• Multi-category agent support — agents can belong to primary and secondary categories
• 4-tier category inference engine — automatically categorizes agents by prompt fingerprint, model name, or agent name
• Unit Economics Intelligence Engine — 6 optimization analyzers: model-switching, provider arbitrage, revenue intelligence, token guardrails, predictive failure, waste detection
• MCP Server with 16 tools across 6 domains (expanded to 23 in v1.8.0)
• Model routing experiments infrastructure
• Outcome defaults for 13 agent categories (Rung 1 of adoption ladder)

Fixed

• Migration numbering collisions between concurrent PRs resolved

[1.5.0] - 2026-02-28

Dashboard & Analytics Bug Fixes

10 display and logic bugs identified during full ICP walkthrough, all fixed and verified.

Fixed

• ROI formula corrected to revenue/cost multiplier (was using profit-margin formula)
• Microcents conversion fixed — costs were inflated 100x
• Duplicate KPI rows on analytics page eliminated
• Individual Calls tab on agent detail page now functional
• Monthly cost enrichment on agent detail (was always $0.00)
• Calls display shows actual count for sub-1000 values (was showing “0k”)
• Agents Used and Calls Used on billing page now access correct API response format
• Missing Budgets link added to Settings hub
• Cross-org session timeline access prevented (security fix)
• Sub-dollar costs display correctly ($0.07 instead of $0)

[1.4.0] - 2026-02-26

Security Audit Remediation

Full security audit identified 10 vulnerabilities — all 10 remediated.

Fixed

• 10 security vulnerabilities remediated across API key management, authentication, rate limiting, CORS, and input validation

[1.3.0] - 2026-02-26

Optimization Engine, Baselines & Anomaly Detection

Added

• 3 new optimization analyzers — Prompt Compression, Batch Eligible, Cost Forecast (6 total)
• Domain relevance matrix — 11 agent templates × 10 suggestion types with weighted scoring for personalized recommendations
• Dual EWMA baseline system — fast (recent behavior) and slow (long-term trends) baselines per agent
• Anomaly detection triggers when metrics deviate beyond configured thresholds
• 5 new onboarding templates — Chatbot, Coding Assistant, Data Pipeline, Content Generator, Research Agent (11 total)
• Daily baseline refresh at 4:00 AM UTC for all active agents

[1.2.0] - 2026-02-24

Pre-GTM Quality & Onboarding Improvements

Added

• Python SDK documentation — auto-instrumentation, streaming, async support
• Langfuse / LangSmith import connector setup instructions
• Integration recommendation engine in onboarding — 2 questions → personalized setup path

Fixed

• Marketing layout TypeScript errors resolved
• API key prefix corrected from al_ to production format

[1.1.0] - 2026-02-24

Integration Pipeline

All 5 data source integration paths verified end-to-end:

Added

• Canonical batch event ingest endpoint with dual SDK format support
• Python SDK streaming support for OpenAI and Anthropic
• Circuit breaker in gateway logger (5 failures → 30s open, auto-recovery)
• Bounded retry and queue in Python SDK transport

[1.0.0] - 2026-02-20

Initial MVP Release

See Integration Guide and API Reference for full details.

Pre-Release (0.1.0 – 0.4.0) — 2026-02-20 to 2026-02-26

Pre-release development included budget system, alert infrastructure, optimization engine, progressive disclosure, security hardening, and IDE-assisted onboarding — all rolled into the 1.0.0 GA release above.

Getting Help

• Documentation: docs.metrxbot.com
• Email: founders@metrxbot.com

Last Updated: 2026-03-13