Coding Ledger By ExecLayer Inc.

Know what
your AI
actually returns.

Coding Ledger reads the receipts already on your machine, Git history, Claude Code and Codex session logs, macOS Screen Time, and turns them into an evidence-backed record of your work and your AI's ROI. Everything runs locally. Nothing leaves your machine.

# install (Python 3.10+, zero dependencies) pipx install coding-ledger # or brew tap BMC-INC/coding-ledger && brew install coding-ledger # run coding-ledger init --author "Your Name,you@example.com" coding-ledger scan --roots "$HOME/Projects" coding-ledger roi

Three questions no other tool answers together

What did your AI cost?

Token usage read straight from local Claude Code and Codex logs, priced at API list rates. API-equivalent value, what it would have cost, not what you paid on a subscription. Unknown models are labeled unpriced, never guessed.

What did it return?

Spend tied to shipped commits and merged PRs: cost per shipped commit, conversion rate per tool, unconverted spend, monthly trends, and your subscription ROI multiple.

How do you actually work?

Human-versus-AI attribution, total involved time verified against Screen Time, streaks, badges, and a shareable builder field report.

Real numbers, reproducible math

AI ROI share card: $4,080 API-equivalent spend, $2.00 per shipped commit across 2,044 commits

An actual Pro share card. Aggregate evidence only: no prompts, no code, no project names.

Hours you can defend

Screen time overlaps git, editor, and steering evidence, so it is an envelope, never blindly added. The formula is printed everywhere the number appears:

screen_hours   = allowlisted foreground seconds / 3600 (18h/day cap)
uncaptured     = max(0, screen_hours - max(human_evidence, shared_ai))
total_involved = max(human_evidence, shared_ai, screen_hours) + independent_ai
               = attributed_total + uncaptured

Uncaptured is the reading, reviewing, and debugging time in coding apps that produced no receipt. It credits to you, never to AI.

Private by construction

The ledger is a local SQLite file. It stores timestamps, counts, and aggregates. It never stores:

prompts or responses source code tool arguments or output credentials or API keys window titles or URLs anything phoned home

Your builder field report, free

The free tier's shareable artifact. Five reproducible dimensions, Steering, Planning, Engineering, Execution, Autonomy, plus an archetype and badges with visible thresholds. No model-generated personality judgment: every number recomputes from your own ledger.

Builder field report card: archetype The Director with dimension scores and badges

Prove how you build without exposing what you build.

Pricing

Free

$0
  • All 11 source scanners and the SQLite ledger
  • roi, today, status, full JSON and markdown reports
  • Offline dashboard with headline ROI totals
  • Public scorecard exports (HTML, PDF, social image)
GitHub →

Pro

$49/year or $99 lifetime
  • Dashboard ROI deep-dive: per-tool, per-model, per-project
  • Monthly trend chart and unconverted-spend view
  • The AI ROI share card export
  • Offline Ed25519 license. No phone-home, ever.
Buy yearly $49 Early lifetime $99

Eleven sources, one ledger

SourceReceipt
GitAuthor-filtered commits, SHA, timestamps, LOC
Claude CodeSession metadata and token usage
CodexSession metadata and token usage
Gemini, Grok, Cursor, AiderSession metadata
VS Code, AntigravityEdit-level activity proxies
GitHubRemote-only commits and LOC, merged PRs
Screen TimePer-app foreground intervals, bundle ids and durations only