token-goat 0.3.1

Hook-based token optimizer for Claude Code on Windows, macOS, Linux, and WSL. Image shrink, re-read dedup, and compaction assist — automatic.

Cuts the tokens Claude Code and Codex CLI burn. Windows, Linux, WSL, and macOS. Install once, then forget it.

---

97.4% image compression  ·  85% smaller reads via surgical CLI  ·  zero ongoing maintenance

---

The problem

Long sessions accumulate waste three ways. Screenshots cross the model at full resolution. A single PNG can land at 3.3 MB. The agent re-reads files it already parsed earlier in the same conversation. And when a session compacts, the summary LLM doesn't know which files were edited or which symbols mattered, so it preserves the wrong things.

Each one is preventable. Token-Goat intercepts all three, automatically.

What changes

Without Token-Goat	With Token-Goat
3.3 MB screenshot lands in model context	84 KB compressed copy — 97.4% smaller
Agent re-reads files from earlier in the session	"Already read this" reminder with narrow slice suggestion
Compaction forgets which files were edited	Structured session manifest injected before compact
Full file read for one function or section	token-goat read file::symbol — about 85% smaller

Four hours of use on the author's machine: 59.7 MB of data that never hit the model, with an estimated 11.5 million tokens avoided.

Install

Windows requirements: Windows 10 or 11 · Python 3.11, 3.12, or 3.13 · uv (winget install astral-sh.uv)

Linux / WSL requirements: Python 3.11, 3.12, or 3.13 · uv (curl -LsSf https://astral.sh/uv/install.sh | sh)

macOS requirements (untested): Python 3.11, 3.12, or 3.13 · uv (curl -LsSf https://astral.sh/uv/install.sh | sh)

uv tool install token-goat
token-goat install

Two commands. Done. Hooks register, a background worker starts at logon and stays out of the way. No terminal popups, no tray icon, no service to babysit.

On Linux and WSL, the worker registers as a systemd user service when systemd is available. On WSL without systemd, and on macOS, the SessionStart hook ensures the worker is running at the start of every Claude Code session.

Codex CLI users

token-goat install --codex

The --codex flag patches both Claude Code and Codex CLI in one pass.

CLI

Command	What it does
token-goat symbol <name>	Jump to a symbol definition
token-goat read "file::symbol"	Pull one function or class, not the whole file
token-goat section "doc.md::Heading"	Pull one Markdown section by heading
token-goat semantic "<query>"	Find code by meaning, not by filename
token-goat map	Get a compact orientation of the repo
token-goat stats	See how many tokens you have saved
token-goat compact-hint --session-id <id>	Inspect the compaction manifest for a session
token-goat doctor	Confirm everything is wired correctly

First token-goat semantic call downloads a small embedding model, about 130 MB, into the token-goat data directory. One-time. Offline after that.

What gets installed?

token-goat install writes the following on your machine — nothing else, anywhere. Every entry is reversed by token-goat uninstall. Run token-goat doctor at any time to see which of these are currently present.

Claude Code integration (~/.claude/)

Path	What
~/.claude/settings.json	Hook entries for SessionStart, PreToolUse (Read, Drive/WebFetch), PostToolUse (Edit/Write/MultiEdit, Read/Grep/Glob), and PreCompact. Plus a Bash(token-goat:*) permission allowlist entry. Existing hooks are preserved; a timestamped .bak is written before any change.
~/.claude/CLAUDE.md	A delimited block (<!-- token-goat-begin --> … <!-- token-goat-end -->) telling the agent to prefer token-goat read / symbol / section over Read / Grep. Any existing content is preserved.
~/.claude/skills/token-goat/SKILL.md	The token-goat skill — the same routing guidance in skill form.

Worker autostart (one of the following, picked by platform)

Platform	Entry
Windows	HKCU\Software\Microsoft\Windows\CurrentVersion\Run\token-goat-worker. No admin rights required.
Linux (with systemd --user)	~/.config/systemd/user/token-goat-worker.service, enabled.
Linux (no systemd, incl. WSL)	~/.config/autostart/token-goat-worker.desktop. On WSL without systemd, the SessionStart hook also starts the worker on every Claude Code session.
macOS (untested)	~/Library/LaunchAgents/com.dfkhelper.token-goat-worker.plist, loaded via launchctl.

The autostart command is pythonw -m token_goat.cli worker --daemon from Token-Goat's uv tool venv. No PyInstaller-style launcher .exe is dropped; AV/EDR products do not behavior-flag this invocation pattern.

Weekly auto-update (Sunday 03:00 local time, runs uv tool upgrade token-goat)

Platform	Entry
Windows	Scheduled task token-goat-update (schtasks).
Linux / macOS	A crontab line tagged with # token-goat-autoupdate.

Data directory (created on first run)

Platform	Path
Windows	%LOCALAPPDATA%\dfk-helper\token-goat\
Linux / WSL	~/.local/share/token-goat/
macOS	~/Library/Application Support/dfk-helper/token-goat/

Contains the symbol index (global.db, per-project .db files), session cache, shrunken-image cache, embedding model (~130 MB, downloaded on the first semantic call), logs, locks, and the dirty-file queue. Nothing outside this directory and ~/.claude/ is written.

With --codex (Codex CLI integration)

Path	What
~/.codex/config.toml	Hooks block with Codex-specific matchers (`view_image
~/.codex/AGENTS.md	A delimited block (<!-- token-goat-codex-begin --> … <!-- token-goat-codex-end -->) with the same routing guidance, adapted for Codex tool names.

Zero maintenance

After install, there is nothing to start, stop, or restart. The worker runs at logon on Windows, Linux, and macOS; on WSL without systemd, the SessionStart hook covers it. Survives reboots on every platform. token-goat uninstall reverses every change, including the startup entry.

Verify

token-goat doctor
token-goat stats

doctor confirms the install is healthy. stats shows cumulative savings.

Security, privacy, and uninstall

No telemetry. No analytics. No background reporting or silent outbound connections.

Outbound network is reserved to three explicit cases:

• First token-goat semantic call downloads the embedding model (~130 MB) into the data directory. Offline after that.
• Google Drive API calls, only if you already authorized Drive in Claude Code. Token-goat never prompts for its own auth.
• Explicit, user-triggered URL fetches via token-goat fetch-image <url>.

Security reports. See SECURITY.md. Email [e-mail]; do not file as a GitHub issue. Reports are acknowledged within 7 days; coordinated disclosure with a 90-day default window.

Windows Defender (optional, Windows only). Real-time scanning slows indexing. To exclude the data folder, open PowerShell as administrator:

Add-MpPreference -ExclusionPath "$env:LOCALAPPDATA\dfk-helper\token-goat"

0x800106ba means the prompt is not elevated; reopen as administrator. On enterprise-managed Windows (domain-joined / Intune), Defender exclusions may be locked by Group Policy. The command will fail; that is expected and harmless.

Uninstall.

token-goat uninstall

Reverses everything in What gets installed?: the scheduled task or systemd unit, the registry value or .desktop or .plist, the hook entries in settings.json, the CLAUDE.md block, the skill directory. Add --codex to also strip the Codex integration. Add --purge to also delete the data directory (cache, index, models, logs). Nothing else on the system depends on it.

About

I built this because long Claude Code and Codex sessions on my machine kept burning context in the same ways: screenshots landing at 2-3 MB, the agent re-reading a file it parsed hours earlier in the same conversation, compactions that forgot which functions were edited. Each felt preventable.

This is a solo project. I use it daily on Windows 11. Tests run across Python 3.11, 3.12, and 3.13.

Available for work

Senior or staff engineering. Developer tools, AI infrastructure, or context management.

I've spent months inside Claude Code's hook system, session management, and compaction pipeline. Not reading the docs. Instrumenting them to see what was actually happening. The work is in this repo.

I build systems that run without babysitting, measure their own impact, and fail quietly. If you're building tooling for developers who work with AI, reach out.

[e-mail]

Disclaimer

Token-Goat runs on your machine and touches your files. The software is provided as-is, without warranty of any kind. DFK Helper LLC is not liable for any damages arising from use. Full terms, including the No Liability clause, are in the LICENSE file.

License

Token-Goat is licensed under the PolyForm Noncommercial License 1.0.0. See the LICENSE file for the full terms.

Individual developers may install and use Token-Goat on their own machines for personal productivity without a commercial license, provided the use does not involve providing Token-Goat as a service to others, incorporating it into a commercial product or platform, or deploying it as shared infrastructure across a team or organization. Employment at a for-profit company does not by itself make use commercial — but if your employer is the primary beneficiary of the deployment, a commercial license applies. When in doubt, email [e-mail].

Commercial use is reserved. That means copying or incorporating this codebase into a product, charging for access to it, or running it as shared infrastructure across a team at a for-profit company. Commercial licensing: [e-mail].

Copyright (c) 2026 DFK Helper LLC.

Patent Pending.