CLI Consumption

CLI Consumption is a local-first Python package for measuring how AI coding CLIs use models, tokens, tools, conversations, and turns. It can analyze one workstation, consolidate copied data from several machines, or send metadata-only snapshots to a central collector.

Codex is fully supported. Claude Code, OpenCode, Kilo Code, and Pi are planned behind the same provider-neutral adapter contract.

The collector deliberately excludes prompts, responses, tool arguments, and credentials. See Privacy before sharing a database or export.

Quick start

The project requires Python 3.14 or newer and uses uv for every development and execution workflow.

From a checkout:

uv sync
uv run cli-consumption collect
uv run cli-consumption export --output reports

Open reports/dashboard.html locally. The generated dashboard is self-contained and makes no network requests.

The dashboard supports time, provider, machine, project, and model filters. It reports period-over-period activity, token composition, cache efficiency, turn latency and duration distributions, technical throughput, context pressure, content-free work-item duration and reliability, configuration cohorts, compactions, subagent delegation, and ingestion quality. Token events are local usage metadata rather than billing records, and a technically completed turn is not a measure of task quality or productivity.

For a dashboard that is safer to share, omit the detailed CSV exports, pseudonymize machine, project, model, and role labels, group tool names, round timestamps to days, and hide small rows in the cohort-comparison view:

uv run cli-consumption export --output shared-report --share-safe

Share-safe reports still disclose aggregate work patterns and must be treated as private operational data.

From PyPI, the CLI can run without a permanent installation:

uv tool run cli-consumption collect
uv tool run cli-consumption export --output reports

To run the latest unreleased GitHub source instead:

uv tool run --from git+https://github.com/Guillaume-Lombardo/cli-consumption \
  cli-consumption providers

Collect from one or more machines

With no source option, collect reads ~/.codex and labels it with the local hostname:

uv run cli-consumption collect --database usage.sqlite

For an offline multi-machine workflow, copy only each machine's Codex sessions/ directory into a trusted analysis location. Do not copy auth.json or other credentials. Then repeat --source:

uv run cli-consumption collect \
  --source desktop=/data/codex/desktop \
  --source laptop=/data/codex/laptop \
  --source server=/data/codex/server \
  --database usage.sqlite

Globally identical conversation IDs are deduplicated. If copies differ, the most complete rollout wins. Explicit project mappings use the longest matching original working-directory prefix:

uv run cli-consumption collect \
  --source desktop=/data/codex/desktop \
  --project cli-consumption=/home/me/dev/cli-consumption

SQLite and PostgreSQL

A file path selects SQLite. A SQLAlchemy URL selects PostgreSQL:

uv run cli-consumption collect --database usage.sqlite
uv run cli-consumption collect \
  --database postgresql+psycopg://usage@localhost/cli_consumption

Pass credentials through environment variables or a secret manager rather than shell history. CLI_CONSUMPTION_DATABASE can supply the database setting.

Central collector API

Offline imports are the simplest choice for personal use and air-gapped machines. A central API is useful for recurring collection across machines.

Start the collector locally:

export CLI_CONSUMPTION_API_TOKEN="$(your-secret-provider)"
uv run cli-consumption serve \
  --database postgresql+psycopg://usage@localhost/cli_consumption \
  --host 0.0.0.0

Send a snapshot from another machine:

export CLI_CONSUMPTION_API_TOKEN="$(your-secret-provider)"
uv run cli-consumption sync --endpoint https://usage.example.test

The application refuses to bind beyond localhost without a token. Production deployments must also place TLS and normal operational controls in front of the ASGI server. See Architecture for trade-offs.

Commands

cli-consumption collect    Collect local or copied provider data into SQL
cli-consumption sync       Collect and send a snapshot to a central API
cli-consumption serve      Run the central collection API
cli-consumption export     Write CSV tables and the HTML dashboard
cli-consumption providers  Show supported and planned providers

Run uv run cli-consumption COMMAND --help for every option.

Development

uv sync --all-groups
uv run pre-commit install
uv run pre-commit run --all-files
uv run ruff format --check .
uv run ruff check .
uv run ty check
uv run pytest --cov --cov-report=term-missing
uv build

Development follows short-lived branches and squash-merged pull requests into a protected main. Read CONTRIBUTING.md and AGENTS.md before changing the project.

License

Licensed under the Apache License 2.0. It provides permissive reuse plus an explicit patent grant and contribution protections appropriate for an extensible developer tool.