# Metrx

> Metrx is the scorecard for your AI workforce. Track what every AI agent costs, identify waste, optimize model selection, and prove ROI — in under 60 seconds via MCP server. Free tier, no credit card required.

## What Metrx Does

- Tracks per-agent, per-model LLM costs in real-time (OpenAI, Anthropic, Google, Cohere, Mistral)
- Generates cost optimization recommendations (model switching, token reduction, provider arbitrage)
- Runs A/B experiments comparing models with statistical significance testing
- Detects cost leaks: idle agents, model overprovisioning, missing caching, retry storms
- Links agent actions to business outcomes (revenue attribution, ROI calculation)
- Provides budget governance with hard/soft spending limits and auto-pause enforcement
- Publishes anonymized LLM cost benchmarks from the Metrx network

## What Metrx Does NOT Do

- Metrx is NOT an LLM gateway or proxy — it observes and analyzes, it does not route API traffic
- Metrx is NOT a prompt management tool — use Langfuse or PromptLayer for prompt versioning
- Metrx is NOT an agent hosting platform — it works with your existing infrastructure
- Metrx does NOT store prompt/completion content — only metadata, cost signals, and performance metrics

## For AI Agents (MCP Server)

Install (stdio): `npx @metrxbot/mcp-server`
Remote (HTTP): `https://metrxbot.com/api/mcp`

23 tools across 10 domains:

### Cost Tracking
- `metrx_get_cost_summary` — Get fleet-wide cost summary with spend, call counts, error rates, agent breakdown
- `metrx_list_agents` — List all agents with status, category, cost metrics
- `metrx_get_agent_detail` — Get detailed cost history and performance for one agent

### Optimization
- `metrx_get_optimization_recommendations` — AI-powered savings recommendations (model switch, token guardrails, arbitrage)
- `metrx_apply_optimization` — Apply one-click optimization fix to an agent
- `metrx_route_model` — Get optimal model recommendation based on task complexity
- `metrx_compare_models` — Compare LLM pricing across providers (works Day 0, no data needed)

### Budget Governance
- `metrx_get_budget_status` — View all budget configurations and spending vs limits
- `metrx_set_budget` — Create/update budget with spending limit and enforcement mode
- `metrx_update_budget_mode` — Change enforcement mode or pause/resume budget

### Alerts & Monitoring
- `metrx_get_alerts` — Get active cost spikes, error rate, budget warnings
- `metrx_acknowledge_alert` — Mark alerts as read
- `metrx_get_failure_predictions` — Predictive failure analysis before problems happen
- `metrx_configure_alert_threshold` — Set thresholds that trigger email, webhook, or auto-pause

### Experiments
- `metrx_create_model_experiment` — Start A/B test comparing two LLM models
- `metrx_get_experiment_results` — Get experiment results with statistical significance
- `metrx_stop_experiment` — Stop experiment, optionally promote winner

### Cost Leak Detection
- `metrx_run_cost_leak_scan` — Comprehensive audit: idle agents, overprovisioning, missing caching, error waste, context bloat, missing budgets, arbitrage opportunities

### Revenue Attribution
- `metrx_attribute_task` — Link agent action to business outcome (revenue, cost saving, efficiency)
- `metrx_get_task_roi` — Calculate ROI for a specific agent
- `metrx_get_attribution_report` — Full attribution report with confidence scores

### Upgrade Analysis
- `metrx_get_upgrade_justification` — ROI report for tier upgrades (Free → Lite/Pro) based on usage patterns

### Audit & Compliance
- `metrx_generate_roi_audit` — Board-ready ROI audit with per-agent breakdown and methodology

## MCP Configuration

```json
{
  "mcpServers": {
    "metrx": {
      "command": "npx",
      "args": ["@metrxbot/mcp-server"],
      "env": { "METRX_API_KEY": "sk_metrx_..." }
    }
  }
}
```

## For Developers (REST API & SDKs)

- REST API: https://metrxbot.com/api/v1/
- OpenTelemetry: Send OTLP traces to https://gateway.metrxbot.com/v1/traces
- Python SDK: `pip install metrx`
- TypeScript SDK: `npm install @metrx/sdk`

## Quick Start

1. Sign up at https://metrxbot.com (free, no credit card)
2. Get API key from Settings → Security
3. Install MCP server or SDK
4. See cost data in under 5 minutes

## Links

- Documentation: https://docs.metrxbot.com
- API Reference: https://docs.metrxbot.com/api-reference
- Agent Quickstart: https://metrxbot.com/agent-quickstart
- MCP Server (npm): https://www.npmjs.com/package/@metrxbot/mcp-server
- Cost Benchmarks: https://metrxbot.com/benchmarks
- Pricing: https://metrxbot.com/pricing
- Full Documentation: https://metrxbot.com/llms-full.txt