Metrx is now live — Start tracking your AI costs for free →

Welcome to Metrx — your AI stack, headed for its best proven configuration. This documentation covers everything you need to understand, integrate with, and deploy Metrx.

Metrx finds cheaper, better configurations for every AI workload, proves them against a randomized holdout, and switches on your say-so — as models and prices change. The platform includes:

  • Gateway (Cloudflare Worker) — High-performance LLM proxy that handles authentication, cost calculation, and event logging
  • Web Dashboard (Next.js) — Real-time insights, team management, billing, and outcome tracking
  • Background Workers (BullMQ) — Asynchronous event processing and data synchronization
  • SDK (@metrxbot/sdk) — Easy integration for your applications
  • MCP Server (@metrxbot/mcp-server) — 23 cost intelligence tools for AI agents via Model Context Protocol
  • Enterprise Stack — Supabase (database), Clerk (auth), Stripe (billing)

Quick Navigation

For API Integration

For Infrastructure

For AI Agents (MCP)

  • MCP Server — 23 tools for cost tracking, optimization, budgets, experiments, and ROI

Additional Resources

  • Changelog — Release notes and version history

Getting Started

1. Get an API Key

Sign up at app.metrxbot.com and generate an API key from your organization settings. Your API key grants access to the Gateway.

2. Choose Your Integration Method

Option A: Use the Gateway (Recommended) The Gateway is a drop-in replacement for your existing LLM API calls. You can proxy through Metrx with minimal code changes:

# Instead of calling OpenAI directly:
curl https://api.openai.com/v1/chat/completions \
  -H "Authorization: Bearer sk_openai_xxxx" \
  ...
 
# Route through Metrx:
curl https://gateway.metrxbot.com/v1/chat/completions \
  -H "Authorization: Bearer sk_live_xxxx" \
  -H "X-Provider-Key: sk_openai_xxxx" \
  ...

Option B: Use the SDK For advanced use cases, install the SDK:

npm install @metrxbot/sdk

3. Track Costs and Outcomes

Once integrated, costs are automatically calculated and tracked. View real-time dashboards and set up webhooks for outcomes.

Pricing

Start free. See metrxbot.com/pricing for current plans and the full feature comparison.

Core Concepts

API Key

Your authentication token for the Gateway. Treat it like a password. Never commit it to version control.

Organization

A workspace containing team members, agents, and billing settings.

Agent

A named LLM-powered system within your organization. Each agent tracks its own costs and outcomes.

Session

A conversation thread or workflow instance. Sessions group related LLM calls together.

Event

A single LLM call (chat completion, message, embedding). Events record input/output tokens, cost, latency, and status.

Outcome

A business result tied to one or more sessions. Examples: “successfully handled customer inquiry”, “correctly identified fraud”, “generated valid code”.

Support

For help, visit:

Security & Privacy

Metrx uses industry-standard security practices:

  • API keys are hashed and cached in Cloudflare KV for sub-millisecond authentication
  • Data is encrypted at rest in Supabase using AES-256
  • Transport is TLS 1.3 only
  • Row-Level Security (RLS) ensures teams only access their data
  • Third-party integrations (OpenAI, Anthropic, etc.) never see your data except your requests

For questions about our security practices, contact security@metrxbot.com.

Ready to integrate? Start with the Integration Guide.

Getting Started