Skip to content

Export format

The Export buttons in the list view download the sessions shown there as CSV or JSON. The file contains the sessions that pass the current filters (search, projects, statuses, "With prompts only", and the list filters for model, branch, source, rating, tag, output, dates and cost), in the current sort order.

The file is named cc-calendar-sessions-YYYY-MM-DD.csv or .json, using the date of the export.

Exports contain your session data

Titles are taken from your prompts, paths come from your machine, and your notes are included. Treat exported files like the transcripts themselves.

Fields

CSV and JSON carry the same fields, in this order.

Field Type Description
id string Claude Code session ID
title string Session title: Claude Code's generated title, else the first line of the first prompt
project string Project name as shown in the app
project_path string Working directory of the session
source string Name of the Claude config directory the session was read from, as in the Source column (local with the default ~/.claude). See Several config directories
branch string or null Git branch recorded in the log
status string running, waiting, done or interrupted
rating string or null How you rated the session: done, partial or failed; null when not rated. See Notes and tags
start string First activity, ISO 8601 with your UTC offset, e.g. 2026-09-29T20:15:00+09:00
end string Last activity, same format
active_minutes number Active time: the length of the drawn bars, one decimal place
span_minutes number or null Minutes from start to end, idle time included, one decimal place
prompts integer Prompts you sent
tokens integer All tokens, subagents included: input, output, cache reads and cache writes
cost_usd number Cost in US dollars, four decimal places
cost_estimated boolean true when the cost is estimated from token usage (shown with ~ in the app)
cache_hit_rate number or null Cache reads as a share of all input-side tokens (input, cache writes and cache reads), from 0 to 1
idle_recache_usd number Estimated cost of rewriting the prompt cache after idle gaps, four decimal places. See Idle re-cache
model string or null Model used for the most requests
effort string or null Effort level most requests ran at: max, xhigh, high, medium or low
claude_code_version string or null Claude Code version recorded in the log (the latest one if it changed)
commits integer Commits made in the session
pull_requests integer Pull requests the session opened or linked to
files_changed integer Files the session edited or wrote
lines_added integer or null Lines added, from Claude Code's cost record (a continued session's own share); null when the session wrote none or its share is unknown
lines_removed integer or null Lines removed, as for lines_added
cost_per_commit number or null cost_usd divided by commits, four decimal places; null without commits
interrupts integer Times you stopped Claude with Esc
api_errors integer API requests that failed (overloaded, rate limited, connection lost and the like)
queued_prompts integer Prompts you sent while Claude was still working, which it read before its turn ended
tool_calls integer Tool calls in the main session; subagents' calls are not counted
tool_errors integer Of tool_calls, those that returned an error, including commands that exited non-zero and tool uses you rejected
friction integer interrupts + api_errors + queued_prompts + tool_errors, as in the Friction column
context_avg integer or null Average context per main-thread request, in tokens: input, cache writes and cache reads. Subagents are not included. null without requests
context_peak integer or null Largest context of a single request, as for context_avg
context_bloated boolean true when at least 20 requests each resent more than 200k tokens. See Context size
tags array of strings Tags you gave the session, in the order shown. In CSV, joined with ; (tags cannot contain commas). Empty when there are none. See Notes and tags
note string Your note on the session; empty when there is none. Line breaks are kept

active_minutes depends on the Split after … idle setting. Idle gaps longer than that setting are not counted.

Costs are rough figures, not billing data. See About costs.

CSV

  • UTF-8 with a byte order mark, so Excel reads non-ASCII titles correctly. Lines end with CRLF.
  • The first row is the header, with the field names above. There are no other metadata rows.
  • Fields containing a comma, a double quote or a line break are quoted, with " doubled (RFC 4180).
  • Empty cells stand for null. Booleans are true / false.
  • Some text values start with =, +, -, @, a tab or a carriage return. These are prefixed with ', so spreadsheets do not run them as formulas.

JSON

JSON wraps the records in an object that identifies the format:

{
  "format": "cc-calendar.sessions",
  "schema_version": 1,
  "generator": "cc-calendar 0.7.1",
  "exported_at": "2026-10-04T10:00:00+09:00",
  "sessions": [
    {
      "id": "0b6c1a52-…",
      "title": "Checkout page redesign",
      "project": "acme-web",
      "project_path": "/work/acme-web",
      "source": "local",
      "branch": "main",
      "status": "done",
      "rating": "partial",
      "start": "2026-09-29T20:15:00+09:00",
      "end": "2026-09-29T23:19:00+09:00",
      "active_minutes": 113,
      "span_minutes": 184,
      "prompts": 4,
      "tokens": 6035399,
      "cost_usd": 3.4354,
      "cost_estimated": true,
      "cache_hit_rate": 0.9712,
      "idle_recache_usd": 0.1825,
      "model": "claude-sonnet-5-5",
      "effort": "high",
      "claude_code_version": "2.1.0",
      "commits": 2,
      "pull_requests": 1,
      "files_changed": 6,
      "lines_added": 182,
      "lines_removed": 40,
      "cost_per_commit": 1.7177,
      "interrupts": 1,
      "api_errors": 0,
      "queued_prompts": 2,
      "tool_calls": 84,
      "tool_errors": 5,
      "friction": 8,
      "context_avg": 104512,
      "context_peak": 166830,
      "context_bloated": false,
      "tags": ["redesign", "PR review"],
      "note": "Waiting for design sign-off"
    }
  ]
}
Key Description
format Always cc-calendar.sessions
schema_version Version of the field definitions above
generator cc-calendar version that wrote the file, for troubleshooting
exported_at When the file was written
sessions The records, in list order

Numbers and booleans keep their JSON types. A missing value is null.

Pull requests

The Export buttons of a Pull requests table (the pane and the project page) download its rows as CSV or JSON, newest PR first. The file is named cc-calendar-pull-requests-YYYY-MM-DD.csv or .json. CSV follows the rules above.

Field Type Description
url string Web URL of the pull request
number integer or null PR number
repository string or null owner/name
title string or null Title given to gh pr create; null when it was not recorded (for example when the PR was opened another way)
head_branch string or null Branch the PR was opened from
created string When the PR was opened, ISO 8601 with your UTC offset
opened_in string ID of the session that opened it
sessions array Sessions that worked on it, the opener first. In CSV, their IDs joined with ;. In JSON, objects with id and cost_usd (that session's share)
active_minutes number Active time spent on it over all its sessions, one decimal place
requests integer API requests made for it, subagents included
tokens integer Tokens of those requests
cost_usd number Cost of those requests in US dollars, four decimal places
cost_estimated boolean true when any part of the cost is estimated from token usage
commits integer Commits made for it

The totals cover every session that worked on the PR, also sessions outside the displayed range or hidden by the filters. How work is attributed to a PR is described in Features; active_minutes depends on the Split after … idle setting.

The JSON file has "format": "cc-calendar.pull_requests", schema_version, generator and exported_at as above, the records under pull_requests, and the work of the table's sessions that went to no PR under unattributed:

{
  "format": "cc-calendar.pull_requests",
  "schema_version": 1,
  "generator": "cc-calendar 0.7.1",
  "exported_at": "2026-10-04T10:00:00+09:00",
  "pull_requests": [
    {
      "url": "https://github.com/acme/acme-web/pull/42",
      "number": 42,
      "repository": "acme/acme-web",
      "title": "Redesign the checkout page",
      "head_branch": "checkout-redesign",
      "created": "2026-09-29T23:10:00+09:00",
      "opened_in": "0b6c1a52-…",
      "sessions": [
        { "id": "0b6c1a52-…", "cost_usd": 3.1021 },
        { "id": "5d0e77c4-…", "cost_usd": 0.8740 }
      ],
      "active_minutes": 131.5,
      "requests": 212,
      "tokens": 7120944,
      "cost_usd": 3.9761,
      "cost_estimated": true,
      "commits": 3
    }
  ],
  "unattributed": {
    "sessions": 1,
    "active_minutes": 12,
    "requests": 18,
    "tokens": 402311,
    "cost_usd": 0.3333,
    "cost_estimated": true,
    "commits": 0
  }
}

Versioning

Each format (cc-calendar.sessions, cc-calendar.pull_requests) has its own schema_version. It is bumped when a field is renamed or removed, or when its meaning or unit changes. Adding a field does not bump it, so readers should ignore fields they do not know.

CSV has no version field. Read columns by their header names, not by position.

Schema version cc-calendar Changes
1 0.3.0 First version
1 0.4.0 Added source, tags and note
1 0.6.0 Added pull_requests, files_changed, lines_added, lines_removed, cost_per_commit, rating, idle_recache_usd, and interrupts, api_errors, queued_prompts, tool_calls, tool_errors and friction, and context_avg, context_peak and context_bloated. Added the pull request export (cc-calendar.pull_requests, schema version 1)