Skip to main content

​
Dashboard

Usage data is recorded each time a request is routed through manifest/auto. The main dashboard shows:
  • Total messages — number of LLM calls.
  • Total tokens — input, output, and cache read tokens.
  • Total cost — calculated from model pricing and token counts.
  • Messages over time — LLM calls charted by period.
  • Cost over time — spend charted by period.
  • Model distribution — which models handled what percentage of requests.

​
Metrics captured

MetricDescription
MessagesNumber of LLM calls
Input tokensPrompt tokens sent
Output tokensCompletion tokens received
Cache read tokensTokens served from cache
Cost (USD)Calculated from model pricing × tokens
Duration (ms)Round-trip latency
ModelWhich LLM handled the request
Routing tiersimple / standard / complex / reasoning
Agent nameThe OpenClaw agent that sent the request

​
How cost is calculated

Manifest has pricing data for 300+ models across Anthropic, OpenAI, Google, DeepSeek, and others. 
Cost = input_tokens × input_price + output_tokens × output_price
Pricing updates in the background. In local mode, it syncs from OpenRouter.

​
Message log

Every LLM call is recorded. You can browse the log with:
  • Pagination across all requests.
  • Filters for agent, model, and time range.

​
Data storage

PostgreSQL hosted at app.manifest.build. Data persists across devices and is accessible from any browser.