Recipes

Copy-paste one-liners for common questions. Every example here was run against the shipped binary and the output shape shown is what you'll see.

Two conventions in use throughout:

• Human-readable output on the page — dollars are rounded to two decimals, turn durations are rendered in seconds, and SQL queries use sqlite3 .mode box for a real table.
• Pipe-friendly raw values are called out where relevant — they're single-number jq queries you can embed in scripts without post-processing.

Spend

Grand total (formatted)

claudex summary --json \
  | jq '.total_cost_usd' \
  | awk '{printf "$%\047.2f\n", $1}'

$33,187.40

(Raw number for piping: claudex summary --json | jq '.total_cost_usd'.)

Cost this week

claudex summary --json \
  | jq '.cost_this_week_usd' \
  | awk '{printf "$%\047.2f\n", $1}'

$7,847.78

Top 10 projects by cost

claudex cost --limit 10

The text table already formats dollars and token counts — no jq needed.

Which Opus variant cost the most?

claudex models --json \
  | jq '[.[] | select(.model_family == "Opus")]
        | sort_by(-.cost_usd)[0]
        | {model, sessions: .session_count, cost_usd: (.cost_usd * 100 | round / 100)}'

{
  "model": "claude-opus-4-6",
  "sessions": 467,
  "cost_usd": 30051.13
}

Sessions

Most recent five

claudex sessions --limit 5

All sessions in one project (JSON)

claudex sessions --project claudex --json --limit 100

Sessions that touched a specific file

claudex sessions --file src/index.rs --limit 10

Search

Where did I first discuss a topic

claudex search "foreign key" --limit 1

Case-sensitive search in one project

claudex search CamelCaseThing --project my-app --case-sensitive

Turns

The index stores turn durations in milliseconds. The recipes below convert to seconds so the output is legible at a glance.

One-line timing profile for a project

claudex turns --project claudex --json \
  | jq -r '.[0] | "p50: \(.p50_duration_ms/1000 | round)s   p95: \(.p95_duration_ms/1000 | round)s   max: \(.max_duration_ms/1000 | round)s   (n=\(.turn_count))"'

p50: 66s   p95: 572s   max: 657s   (n=26)

Projects with the slowest average turn

claudex turns --json \
  | jq -r 'sort_by(-.avg_duration_ms)[:5][]
           | "\(.avg_duration_ms/1000 | round | tostring + "s") \t \(.project)"'

2838s    /Users/you/projects/comfyui-nix
373s     /Users/you/projects/quantierra-ui
324s     /Users/you/projects/mold
321s     /Users/you/projects/site
316s     /Users/you/projects/nixos

Files

Top 10 most-edited files

claudex files --limit 10

Files with 20+ edit events

claudex files --json | jq -r '.[] | select(.modification_count >= 20) | .file_path'

Files touched by the most distinct sessions

claudex files --json includes distinct_session_count, so you no longer need raw SQL for this:

claudex files --json \
  | jq -r '.[] | select(.distinct_session_count >= 5) | [.file_path, .distinct_session_count] | @tsv'

Session drill-down

Inspect one session

claudex session <session-id-prefix>

Pull just the files, tools, and stop reasons

claudex session <session-id-prefix> --json \
  | jq '{files: .files_modified, tools: .tools, stop_reasons: .stop_reasons}'

Export

Export every session in a project to a directory

mkdir -p exports
claudex sessions --project claudex --json --limit 1000 \
  | jq -r '.[].session_id' \
  | while read sid; do
      claudex export "$sid" --output "exports/$sid.md"
    done

Export a session as JSON and count messages per role

claudex export <session-id-prefix> --format json \
  | jq '[.messages[].message.role] | group_by(.) | map({role: .[0], count: length})'

[
  { "role": "assistant", "count": 9 },
  { "role": "user", "count": 5 }
]

PRs

claudex prs already deduplicates by pr_url — you get one row per unique PR, with the most recent mention's timestamp.

PRs this month

claudex prs --json \
  | jq -r --arg m "$(date +%Y-%m)" \
       '.[] | select(.timestamp | startswith($m))
              | "#\(.pr_number)  \(.pr_repository)  \(.pr_url)"'

#14  utensils/claudex  https://github.com/utensils/claudex/pull/14
#13  utensils/claudex  https://github.com/utensils/claudex/pull/13
#12  utensils/claudex  https://github.com/utensils/claudex/pull/12

PR count per project

claudex prs --json \
  | jq 'group_by(.project)
        | map({project: .[0].project, prs: length})
        | sort_by(-.prs)'

Summary

Markdown block for standup notes

COST=$(claudex summary --json | jq '.cost_this_week_usd' \
         | awk '{printf "%\047.2f", $1}')
claudex summary --json | jq -r --arg cost "$COST" '
  "## Claude Code this week\n\n" +
  "- Sessions: \(.sessions_this_week)\n" +
  "- Cost:     $\($cost)\n" +
  "- Tokens:   \(.total_tokens / 1e9 * 100 | round / 100)B\n"'

## Claude Code this week

- Sessions: 286
- Cost:     $7,847.78
- Tokens:   18.01B

Paste straight into your standup notes.