What should we monitor for coding-agent usage?

Track model, provider route, token count, latency, status, user or team key, and daily spend. Coding agents can create large bursts of token usage, so separate keys and limits make reviews easier.

Claude Code LLM Gateway Integration

Agent

Claude Code

Traffic grouped under one Universal API key.

Owner

Engineering

Team-level spend attribution.

Logs

Request-level

Debug status, latency, and model usage.

Integration basics

Frequently asked questions

Can finance see Claude Code spend separately from app traffic?

Yes. Give Claude Code its own Universal API key or project key. Concentrate can then separate coding-agent spend from product traffic by key, team, model, provider, and time window.

Can we use Claude Code without handing separate provider accounts to every engineer?

LLM Gateway

Teams

Integrations

Platform

Legal

LLM Gateway

Teams

Integrations

Platform

Legal

Use Claude Code with model access and spend tracking

What your team gains with Concentrate

Track agent spend

Keep access scoped

Switch model routes

Frequently asked questions