OpenTab

Your AI coding tools keep a tab. OpenTab opens it.

_{Anonymized demo data — click the reel for the full-quality video.}

_{One reel, every view — trends, a calendar spend heatmap, drill from a month down to a single session, and live theming}

_{▶ The full tour on YouTube — every view, walked through}

_{Also a web browser — opentab --web renders the same data as one self-contained, shareable page}

A terminal browser for your AI coding spend. It reads the records your tools already keep on disk — a dozen of them, from OpenCode and Claude Code to Codex and Copilot — and shows where the tokens and the money went: by month, day, project, session, and model, all the way down the subagent tree. One tool at a time, or all of them merged.

Plenty of things will print you a token total. This one is built to be poked at: drill in, filter as you type, rescope the dates, sort, keyboard or mouse. And it's just a reader — no backend, no telemetry, no account, every file opened read-only. The runtime is the Python standard library, so pipx install opentab-ai pulls in nothing else.

Install

Python 3.9+ and a terminal — macOS, Linux, WSL, and native Windows. The PyPI package is opentab-ai; the command it installs is opentab.

pipx install opentab-ai                     # recommended
brew install hamidi-dev/tap/opentab         # Homebrew (macOS / Linux)
pip install --user opentab-ai               # pip
curl -fsSL https://raw.githubusercontent.com/hamidi-dev/opentab/main/install.sh | bash

Upgrade with pipx upgrade opentab-ai, brew upgrade opentab, or pip install -U --user opentab-ai.

Usage

opentab                          # the browser, all time
opentab web                      # the same browser, in your web browser
opentab --demo                   # anonymized — safe for screenshots and live demos
opentab doctor                   # what's found, what isn't, and why
opentab pull laptop workstation  # every machine, merged into one tab

? shows the full keymap in-app; it's also in docs/keys.md, along with the rest of the reference in docs/.

What you get

• Drill, don't scroll — month → day → project → session → model, down the recursive subagent tree, with a live fzf-style filter and live date-range scoping.
• Trends — daily / weekly / monthly charts, a calendar spend heatmap, and model / provider / harness rankings. Every one of them navigable down to a single session.
• Inside a session — Turns: what each prompt cost, in order, with the tool calls it made. Tools: token attribution per tool and MCP server, over a treemap where area is total spend and shade is cost per call. Context: the window filling up over time, compaction markers included, plus an estimate of what filled it.
• Honest money — subscription usage shows its true $0; $ reprices it at API list rates and P shows the per-model table behind the estimate. w arms a what-if model: what if the expensive one had done the subagents' work too?
• Every machine, one tab — opentab pull gathers each box's spend over SSH and merges it into one browser you can filter by machine.
• A web twin — opentab web serves the same browser live, --html writes it as one self-contained file: same keys, same overlays, every view a shareable deep link. docs/web.md
• Yours to drive — lazygit-style keyboard and mouse, 30 bundled themes shared by the TUI and the web page, every key remappable, CSV export of any view, and your harness, range, sort and $ view remembered between runs.
• Private by construction — local-only, read-only, no telemetry, no accounts.

The tools it reads

OpenCode · Claude Code · Codex CLI · GitHub Copilot (its CLI and Copilot Chat in VS Code) · pi-agent · omp · OpenClaw · zaly · Hermes · and CSV/JSONL logs of your own API requests.

Point it at nothing and --harness auto merges every tool it finds; --harness NAME narrows to one, H switches live. You can also just hand it a file — opentab requests.csv, opentab path/to/opencode.db — and the harness is inferred.

docs/sources.md has the rest: where every tool's records live, its flags and env vars, how its cost is derived, the quirks (Copilot's opt-in export, Codex's cumulative counters, the metered-vs-subscription split), the CSV/JSONL schema you can write against, and the merged view.

Every machine, one tab

Code on more than one box? opentab pull gathers each machine's spend over SSH — all in parallel — and opens them merged into one browser, filterable and drillable by machine.

opentab pull laptop workstation gpu-box   # fetch all three, open the fleet
opentab pull                              # later: refresh every machine you've saved

No agent, nothing to install, nothing listening — the remote only needs opentab on its PATH.

Live prices in your sidebar

herdr-opentab puts what each running agent has spent right next to it in the Herdr sidebar — that pane's own session, subagents included, kept up to date while it works.

_{Amounts synthetic, everything else real.}

About the money

A session that shows tokens against $0.00 isn't a bug: usage was recorded without a per-token price, which is what a subscription or credit plan looks like from the outside. That money isn't missing, it's billed elsewhere — so OpenTab reports unpriced tokens rather than guessing. Press $ and it reprices them at published API list rates, from a models.dev snapshot bundled with each release (nothing is fetched at runtime); P shows the rates behind the estimate. docs/pricing.md

Something looks off?

opentab doctor

Which copy of OpenTab is talking, every harness (found — or not found and why, with the fix), your terminal's colour and glyph support, the price catalog, its own files. It reports and never repairs, and it reads no transcript, so the output is safe to paste into an issue as-is. docs/troubleshooting.md

FAQ

Development

CI runs Ruff, unit tests, and ShellCheck. CONTRIBUTING.md has the local setup, test/lint commands, hooks and commit conventions; docs/architecture.md has how the code is put together.

License

MIT — see LICENSE.