promptcmd

Appearance

Monitoring Usage ​

Track API usage, token consumption, and performance metrics across all your prompts and providers using promptctl stats.

View All Statistics ​

Track usage across all providers and models:

bash
$ promptctl stats

provider      model                     runs     prompt tokens     completion tokens     avg tps
anthropic     claude-opus-4-5           15       1988              1562                  31
openai        gpt-5-mini-2025-08-07     2        88                380                   42

This shows:

  • provider: The LLM provider used
  • model: The specific model invoked
  • runs: Number of times the prompt was executed
  • prompt tokens: Total input tokens consumed
  • completion tokens: Total output tokens generated
  • avg tps: Average tokens per second (throughput, indicating generation speed)

INFO

Statistics are stored in ~/.promptcmd/stats.db. See Lookup Paths for more details on file locations.

View Last Execution ​

View statistics for the most recent prompt execution:

bash
$ promptctl stats --last

provider      model               prompt tokens     completion tokens     time
anthropic     claude-opus-4-5     206               168                   5

The time column shows the execution duration in seconds. This is useful for debugging or understanding the token usage and performance of a specific prompt run.

Dry Run Testing ​

You can test prompts using promptctl run --dry to preview what will be sent without calling the API:

bash
$ promptctl run --dry PROMPTNAME -- [PROMPT ARGS]

This lets you validate your prompt template and arguments without consuming tokens or incurring API costs. See Executing Prompts for more details on dry run mode with other execution methods.

Caching ​

See Caching