Chisle
The coding-agent plugin that bills 52% of a bare model across 20 live tasks. Terse persona + tool-output compressor + context diet.
What a branch or pull request cost in AI-agent tokens.
Translation pending
Developer toolsGitHub @IAmSkweetisOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
Repository information has not been collected yet.
No updates detected yet. GitHub releases will appear here.
Report what a branch or PR cost in AI-agent tokens, including subagent and worktree work. Reads Claude Code transcripts locally.
Loading the full README.
Read the full README ↗branch-tally is a utility to help you track AI costs across multiple workspaces, agents and
worktrees. Claude Code is the only harness supported today. It reads local transcripts and assigns
calls to a record in a ledger. Persisting the ledger is opt-in: branch-tally record writes only
in a repository that has a .tally/ directory, or to a file named with --ledger.
The ledger only stores relevant metadata from the transcript. Prompts and messages are never stored.
As I've been working on different projects, I like to keep an eye on the costs of a PR and a project. There are not any good tools for doing this. I had found different plugins that get you partly there. The different labs do not provide the best metrics or tools for this either.
What I had found are different projects that will give you a running token cost in a terminal. Or a plugin or extension for different harnesses that are specific to a branch. But nothing really fit how I think most folks are using AI in their workflows. And there is definitely a gap on a larger project calculation.
So I had created a python script and added some instructions to a skill to include the usage costs into my PRs.
The goal with branch-tally is to expand on this idea. And honestly, fix some bugs that I've been putting off, too. We'll see where we get.
branch-tally records its own development. This is to both provide an example and to be transparent about the development costs of the project.
The way I use branch-tally is I also add a report at the end of each PR.
You can read more about this setup in the Workflow Examples in the documentation.
Linux (x86_64, aarch64) and macOS (Intel, Apple silicon):
curl -fsSL https://github.com/IAmSkweetis/branch-tally/releases/latest/download/install.sh | sh
Windows (PowerShell):
irm https://github.com/IAmSkweetis/branch-tally/releases/latest/download/install.ps1 | iex
The binary goes to ~/.local/bin (%USERPROFILE%\.local\bin on Windows), after the archive is
checked against the release's SHA256SUMS. The installers never edit your PATH; they print the
line to add. Their options, verifying a download by hand and building from source are in the
manual's Installation chapter.
To move to a newer release later, run branch-tally update. It checks the download against
SHA256SUMS and asks before replacing the binary.
Then, inside a repository where you have worked with Claude Code:
branch-tally report # the current branch's token usage and estimated cost
branch-tally report --pr # write the table into the branch's pull request
Getting Started walks through the first report and how to read it.
Estimated costs come from a price table sourced from LiteLLM.
The binary bundles a dated snapshot, and every report states its date. branch-tally prices update
caches newer prices, which take precedence, and a repository can override them. The manual's
prices reference provides details.