Your agents keep a tab. Open it.

Your agent ran for an hour.
What actually happened?

Follow tokens, cost and context down to the prompts, subagents and tool calls behind them. Across Claude Code, Codex, Copilot and more.

Local & read-only · No account · No telemetry

A number → the work behind itSynthetic demo
A year of history. One session. Every step of the work.
Open the example in OpenTab →

Your history is already there.OpenTab reads the records your coding tools keep.Try the real terminal or browser app below.

pipx install opentab-ai
then opentab in a terminal, or opentab web for the browser

Now investigate it yourself.

Run the real OpenTab right here, nothing to install, with a year of synthetic history. Pick a session and follow the work behind the numbers.

~ opentab

Runs locally in this tab. Uses synthetic history. Python runtime and demo download once; conversations load on demand.

Token economics

Tokens tell one story. Dollars tell another.

A token isn't a unit of cost. Pick a type to see how its share of tokens compares with its share of estimated cost.

Loading the synthetic showcase breakdown…

Synthetic history · All tokens priced at API list rates, rather than mixing recorded spend and estimates. How pricing works →

Keep following the evidence

What filled the window? What changed?

OpenTab Context view showing context growth and an estimated composition breakdown.
See the context grow, and where compaction happened. Follow the recorded usage curve, with an estimate of how prompts and tool calls filled the window.
Explore session views →

Across your coding tools

Reads the tools you already use.

OpenTab finds their local records on its own and opens them read-only. What you can drill into depends on what each tool writes down.

See which details each harness provides →

Fits your workflow.

Terminal & browser

Investigate where you work: keyboard-driven in the terminal, or in a browser served on localhost.

opentab
opentab web

Across machines

Bring your laptop and workstation into one view over SSH. Pull usage summaries; fetch individual remote turns when you open them. No background agent needed.

opentab pull laptop workstation
Connect your machines →

Share a report

Export a self-contained HTML report to open offline or share. Optional --demo scrambling lets you disguise selected fields; dates and the shape of the data stay real.

opentab --html report.html
What an export contains →

Local & read-only

No account. No telemetry. Your tools’ databases and transcripts stay unchanged. OpenTab stores its own notes, preferences and caches separately.

Read the privacy boundaries →

Settle the tab.

pipx install opentab-ai
Python 3.9+ · macOS, Linux, WSL and Windows · standard library only