OpenRouter Activity and API keys

OpenRouter Activity is billing truth for gateway traffic: model slug, cost in credits, API key, and cache fields per generation.

Hub: Tokenminning with OpenRouter.

This page answers openrouter activity, openrouter token usage, and openrouter cost tracking.

Open the dashboard

1. Go to openrouter.ai/activity 
2. Filter by API key, model, or date range
3. Click a row’s detail view for token breakdown including cache metrics

Export CSV or PDF for weekly reviews — group by Model or API Key to find dominant spend.

Per-key isolation

Rotate keys when a tool leaks into the wrong environment. Queries like openrouter_token ext:ini usually mean hunting leaked keys in repos — use env vars and secret scanners, not committed config.

Analytics API

For programmatic spend reviews, use the Analytics API  with a management key :

curl "https://openrouter.ai/api/v1/analytics/query" \
  -H "Authorization: Bearer $OPENROUTER_MANAGEMENT_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "metrics": ["cost", "prompt_tokens", "completion_tokens"],
    "dimensions": ["model"],
    "filters": [],
    "start_time": "2026-01-01T00:00:00Z",
    "end_time": "2026-01-31T23:59:59Z"
  }'

Pass workspace_id to scope results to one workspace. Use group_by=workspace to split rows per project.

Reconcile with client telemetry

After a heavy week: which key and model slug moved most? Split keys if one agent consumed the pool.

Organization accounts

Shared credit pool — Activity shows all members’ metadata (model, cost, timing) but not prompt text. Per-developer keys improve cost attribution.

Spending guardrails

• Set spending limit on each key at creation
• Org-level alerts before credits deplete — notifications 
• Block CI keys from frontier slugs in client allowlists
• Weekly CSV export for cost attribution by model and key

Troubleshooting

Activity shows frontier model, client shows haiku — fallback succeeded on a rate limit. Activity records the model that actually ran.

Cost higher than token count suggests — check cache_write_tokens (first-turn cache establishment) and platform fee on small requests.

No data for yesterday — Activity covers completed UTC days. Allow a few hours lag for final aggregation.

Related

• OpenRouter hub
• Guardrails
• Prompt caching — inspect cache in Activity detail
• Model routing
• Usage accounting