Topics / #token-usage

Token-usage projects on GitHub

Tracking and metering LLM token consumption and costs.

As of , the most-starred Token-usage projects on GitHub are caveman (113.1K stars), headroom (75.5K stars) and 9router (30.9K stars). GitHub.Locker tracks 48 Token-usage repositories with 423.2K stars combined. The fastest-growing on October 16 was caveman, with +330 stars in a day.

48Projects
423.2KStars in total
+1.8KStars in a day
Browse in the catalog →github.com/topics/token-usage ↗

Hot right now

Most stars gained in a day · Oct 16

+330

caveman

JuliusBrussee/caveman · ★ 113.1K

Caveman is a skill plus an optional local proxy for AI coding agents that makes them answer in short, caveman-style prose instead of padded explanations, cutting roughly 65% of output tokens and up to 33% of input tokens in the project's own benchmarks without changing the technical content of the answer.

+265

dsh-ops

T-Auto/dsh-ops · ★ 2.6K

dsh-ops is a Windows plugin for the DSH AI-assistant app that bundles a standalone bash, PowerShell 7, and faster Rust-based file tools so the AI agent stops fighting the native PowerShell shell, which cuts wasted tokens, time, and model attention.

+206

cost-xray

tigerless-labs/cost-xray · ★ 5.4K

cost-xray is a local tool that captures the raw API traffic between Claude Code or Codex and their providers, then breaks down exactly what's in each request and what each part costs — system prompts, tool schemas, MCP blocks, cached vs fresh tokens, and more.

+144

billion-context

ranxianglei/billion-context · ★ 1.5K

billion-context is a context-compression proxy that sits between any coding agent and its model API, rewriting Anthropic/OpenAI streams so a 100K context window is enough and a single session can run for months across billions of tokens.

The all-time top

48 projects · by stars

  1. #1 · JuliusBrussee/caveman

    caveman

    Caveman is a skill plus an optional local proxy for AI coding agents that makes them answer in short, caveman-style prose instead of padded explanations, cutting roughly 65% of output tokens and up to 33% of input tokens in the project's own benchmarks without changing the technical content of the answer.

    113.1K
  2. #2 · headroomlabs-ai/headroom

    headroom

    Headroom compresses everything an AI agent reads - tool outputs, logs, RAG chunks, files, and conversation history - before it reaches the LLM, cutting tokens while keeping the same answers. It ships as a library, a local proxy, and an MCP server.

    75.5K
  3. #3 · decolua/9router

    9router

    9Router is a local AI router and token saver: it connects coding tools like Claude Code, Codex, Cursor and Cline to 40+ providers and 100+ models, auto-falls back from your subscription to cheap and free models, and compresses tool output to save 20-40% of tokens.

    30.9K
  1. 4

    toon

    25.5Ktoon-format/toon · TypeScript

    TOON (Token-Oriented Object Notation) is a compact, human-readable way to serialize JSON data so it spends fewer tokens in LLM prompts. The repo ships a TypeScript SDK, a CLI, and benchmarks.

  2. 5

    CodexBar

    22.6Ksteipete/CodexBar · Swift

    CodexBar is a tiny macOS 14+ menu bar app that keeps AI coding-provider usage limits visible and shows when each usage window resets, so you can plan long tasks instead of guessing whether you have quota left.

  3. 6

    ccusage

    19Kccusage/ccusage · Rust

    ccusage is a command-line tool that reads local usage data from coding agent CLIs and turns it into daily, weekly, monthly, and session reports of token usage and cost.

  4. 7

    RagaAI-Catalyst

    16.2Kraga-ai-hub/RagaAI-Catalyst · Python

    RagaAI Catalyst is a platform for managing and optimizing LLM projects — tracing, evaluating and safeguarding LLM and agent applications, shipped as a Python SDK.

  5. 8

    gitingest

    15.9Kcoderamp-labs/gitingest · Python

    Gitingest turns any Git repository into a prompt-friendly text digest for LLMs. You can also just swap 'hub' for 'ingest' in a GitHub URL (e.g. github.com -> gitingest.com) to get the digest instantly.

  6. 9

    codeburn

    11.5Kgetagentseal/codeburn · TypeScript

    CodeBurn is a free, open-source, local-first tool that tracks AI coding token usage and cost across 41 tools and agents — Claude Code, Cursor, Codex, Gemini, Grok and more — broken down by model, project, and task.

  7. 10

    claude-engineer

    11.2KDoriandarko/claude-engineer · Python

    Claude Engineer is a self-improving AI assistant framework built on Claude 3.5 Sonnet that can generate and manage its own tools during a conversation, expanding its own capabilities over time. It ships with both a CLI and a web interface for software development tasks.

  8. 11

    fast-jev-compaction

    7.7Ktamaratran/fast-jev-compaction · TypeScript

    A Claude Code plugin, also usable as an npm library, that replaces context compaction's summary with Jev decisions: every tool call and result is scored in fast requests, stale ones are dropped or truncated, and everything kept stays verbatim.

  9. 12

    pxpipe

    7.5Kteamchong/pxpipe · TypeScript

    pxpipe is a local proxy that cuts Claude Code's (and other LLM clients') input token usage by rendering bulky context — system prompts, tool docs, and conversation history — as compact PNG images instead of raw text, since an image's token cost depends on its pixel size, not the amount of text packed into it.

  10. 13

    claude-token-efficient

    6.1Kdrona23/claude-token-efficient · Python

    A single CLAUDE.md instruction file that you drop into a project to make Claude Code's responses terser, cutting output tokens on heavy, repeated workflows.

  11. 14

    tokscale

    5.7Kjunhoyeo/tokscale · Rust

    Tokscale is a terminal tool that tracks how many tokens and how much money you burn across AI coding agents like Claude Code, Codex, Cursor, and Gemini CLI — with a dashboard and a global leaderboard on top.

  12. 15

    cost-xray

    5.4Ktigerless-labs/cost-xray · Python

    cost-xray is a local tool that captures the raw API traffic between Claude Code or Codex and their providers, then breaks down exactly what's in each request and what each part costs — system prompts, tool schemas, MCP blocks, cached vs fresh tokens, and more.

  13. 16

    DeepSeek-Balance-Whale-Widget

    5KMeteorNOX/DeepSeek-Balance-Whale-Widget · JavaScript

    A tiny whale mascot widget that sits in the bottom-right corner of the DSH (DeepSeek Harness) web interface, showing your DeepSeek API balance, today's spending, and a popup after every chat turn showing how much that turn cost.

  14. 17

    PilotDeck

    4.5KOpenBMB/PilotDeck · TypeScript

    An open-source agent operating system for task-oriented, multi-project productivity work, where AI agents run real projects in isolated WorkSpaces. Jointly developed by Tsinghua THUNLP, ModelBest, OpenBMB and AI9Stars.

  15. 18

    openusage

    4.4Krobinebers/openusage · Swift

    OpenUsage is a free, open-source native macOS menu bar app written in Swift that tracks how much of your AI coding subscriptions (Claude, Codex, Copilot, Cursor and more) you've used — session and weekly limits, credits, and spend — all in one popover.

  16. 19

    opencode-dynamic-context-pruning

    4.4KTarquinen/opencode-dynamic-context-pruning · TypeScript

    Dynamic Context Pruning (DCP) is a plugin for OpenCode that automatically cuts token usage by intelligently managing conversation context — compressing stale history into summaries, deduplicating repeated tool calls, and purging failed tool-call inputs.

  17. 20

    Codex-X

    4.2Kyynxxxxx/Codex-X · Rust

    Codex-X is a cross-platform desktop app that gives OpenAI Codex (desktop app and CLI) a visual control panel: it centralizes prompt injection templates, provider/API switching, session syncing, Skills/MCP management, and TOML config editing so you don't have to hand-edit config files.

  18. 21

    claude-devtools

    4Kmatt1398/claude-devtools · TypeScript

    claude-devtools is a free, open-source debugging tool for Claude Code that reads the session logs already saved on your machine and reconstructs tool calls, thinking steps, subagent trees, token usage, and context window breakdowns in a visual desktop UI.

  19. 22

    token-monitor

    3.1KJavis603/token-monitor · JavaScript

    A desktop widget that shows live token usage and AI Tool Limits across 44+ AI coding tools - Claude Code, Codex, Cursor, GitHub Copilot, Cherry Studio, and more - with real-time multi-device sync, historical usage trends, and breakdowns by tool, device, model, session, or project.

  20. 23

    dsh-ops

    2.6KT-Auto/dsh-ops · JavaScript

    dsh-ops is a Windows plugin for the DSH AI-assistant app that bundles a standalone bash, PowerShell 7, and faster Rust-based file tools so the AI agent stops fighting the native PowerShell shell, which cuts wasted tokens, time, and model attention.

  21. 24

    dsh-context

    2.2Kbowenliang123/dsh-context · TypeScript

    dsh-context is a plugin for DeepSeek Harness (dsh) that gives full visibility into an AI agent's context window — a dashboard tab, a sidebar panel, and a /context slash command showing token stats, context composition, trends, compaction/prune events, file activity, and the agent family tree.

  22. 25

    DeepSeek-Bot

    2.1KFei2411/DeepSeek-Bot · JavaScript

    DS Bot is a plugin for DeepSeek Harness (DSH) that turns DSH into a set of long-lived bots you create and talk to, including a main bot, bot-to-bot messaging, and group chats.

  23. 26

    codex-astra-luna-orchestrator

    1.7Kdonvito/codex-astra-luna-orchestrator · Python

    This is a configurable Codex setup that makes GPT-6 Astra the root orchestrator and independent reviewer while GPT-5.6 Luna runs as the default pinned model for execution subagents (explorer, worker, tester, researcher).

  24. 27

    billion-context

    1.5Kranxianglei/billion-context · TypeScript

    billion-context is a context-compression proxy that sits between any coding agent and its model API, rewriting Anthropic/OpenAI streams so a 100K context window is enough and a single session can run for months across billions of tokens.

  25. 28

    workbuddy-switch

    1.3Kchangexbc/workbuddy-switch · Rust

    A desktop app (and webui/npm alternative) for switching between multiple WorkBuddy, CodeBuddy CLI, and CodeBuddy CN IDE accounts, with credit-expiry monitoring, auto check-in, and token usage statistics.

  26. 29

    hermes-jev-skills

    1.1Kkerpopule/hermes-jev-skills · Python

    Hermes Jev Skills gives an AI agent a cheap, fast 'second brain' for small routine decisions (model routing, memory filtering, skill selection, turn compaction, computer/browser use) so the expensive frontier model only does the writing. The decisions are handled by Jev, TypeSafe's calibrated decision model.

  27. 30

    remoteai

    1.1Kpiperhex/remoteai · TypeScript

    Remote AI (formerly Codex Switch) is a graphical AI programming workbench built around OpenAI Codex. It combines a chat-based coding assistant, project/Git management, and multi-account management in one desktop app, with web and mobile companions.

  28. 31

    ThinkWatch-Lite

    824ThinkWatchProject/ThinkWatch-Lite · TypeScript

    ThinkWatch Lite is a local gateway that sits between AI coding clients (Claude Code, Codex, and others) and their upstream providers, letting you switch providers and models without reconfiguring each client, while tracking request cost and blocking malicious tool calls.

  29. 32

    casbin-gateway

    642apache/casbin-gateway · Go

    An open-source gateway for the AI coding agents on your machine: it watches what every agent does, relays their model traffic through one endpoint covering 44 vendors, and grades whether the API behind each key is what you actually paid for.

  30. 33

    mintocode

    490Spearmintai/mintocode · Python

    mic compiles your codebase once into agent-facing 'cards' (what each function does, its gotchas, its callers, its tests), so that Claude Code, Codex or any MCP agent only reads the few paragraphs relevant to a question instead of exploring the whole repo.

  31. 34

    dsh-cost-meter

    435Han-1413141/dsh-cost-meter · JavaScript

    A plugin for DeepSeek Harness (DSH) that tracks session and daily costs, token usage, budgets, provider account balances, and coding-plan subscription quotas, with a bilingual English/Chinese interface.

  32. 35

    Astra-Ares

    308miuuyy/Astra-Ares · JavaScript

    Astra-Ares is an experimental tool that lets Jev, an AI evaluator, adapt GPT-6's reasoning effort while a Codex task runs, matching thinking depth to each next step to reduce token usage.

  33. 36

    tare

    296kelviq/tare · Python

    tare is a tool that lets you ask Claude Code in plain English where your token usage went, diagnosing why you hit a usage limit by reading the session logs Claude Code already keeps on your machine.

  34. 37

    dsh-agent-team-gui

    288toolclub/dsh-agent-team-gui · TypeScript

    dsh-agent-team-gui is an unofficial plugin for DeepSeek Harness (DSH) that lets you build a reusable multi-agent team: one agent plans the work, another implements it, and a third reviews it, with its own Run Center for tracking progress and Token usage.

  35. 38

    deepseek-harness-desktop

    285chokwinlee/deepseek-harness-desktop · Kotlin

    DSH Desktop is a compact desktop app that runs the official DeepSeek Harness (DSH) coding-agent Web UI and runtime inside a native window, plus it ships native iPhone and Android 'Remote' companion apps to view and steer the same sessions from your phone while all execution stays on the computer.

  36. 39

    pi-posthorse

    272fitchmultz/pi-posthorse · TypeScript

    Posthorse is an extension for the Pi coding agent (specifically the fitchmultz/pi fork) that handles context window rollovers without summarization: when a session nears its limit it starts a fresh window with a durable handoff and a fully recoverable transcript instead of compressing the history into a summary.

  37. 40

    memex

    266nicosuave/memex · Rust

    Memex indexes your local coding-agent conversations (Claude Code, Codex, Cursor, and more) so you can search past work, read original transcripts, and resume sessions in the original tool.

  38. 41

    dsh-damage-pulse

    265wssfk12138/dsh-damage-pulse · JavaScript

    dsh-damage-pulse is a monitoring plugin for DSH (DeepSeek Harness) that tracks DeepSeek API token usage, cost, and account balance, showing per-call billing, session totals and a global balance card, with a whale-girl mascot animating each charge event.

  39. 42

    dsh-plugin-agent-workflow

    255xuanyuanzhifeng/dsh-plugin-agent-workflow · TypeScript

    It is a web UI plugin for DeepSeek Harness (DSH) that adds a new 'Workflow' tab, turning each user conversation turn into a clear execution chain of the agent's model requests, model responses, and tool calls.

  40. 43

    codex-trajectory

    249icesixgod/codex-trajectory · Python

    Codex Trajectory is a privacy-aware Codex plugin whose MCP tools turn local Codex task logs into an event ledger and interactive timeline, showing turns, reasoning summaries, tool timing, subagents, token usage, and failures without touching the original logs.

  41. 44

    underclass

    237ghuntley/underclass · Rust

    It pools multiple ChatGPT/Codex and GitHub Copilot subscriptions behind one OpenAI-compatible endpoint — when one account runs dry it cools down until its window resets and comes back on its own.

  42. 45

    AgentMeasure

    218roy-tong/AgentMeasure · Python

    AgentMeasure is an open measurement layer for AI-agent usage: a free CLI and SDK that audit Codex and Claude Code session logs and verify that agent metrics and metering numbers mean what their labels claim.

  43. 46

    squeez

    217claudioemmanuel/squeez · Rust

    Squeez is a hook-based token compressor for seven AI coding CLI hosts (Claude Code, Copilot CLI, OpenCode, Gemini CLI, Codex CLI, Pi, Hermes) that intercepts tool output and shrinks bash results by up to 95% without losing exact identifiers.

  44. 47

    TokenLedger

    202zh667/TokenLedger · JavaScript

    TokenLedger is a plugin for DeepSeek Harness (dsh) that tracks token usage and attributes it to the relay site that actually served each request, with zero configuration and no credentials needed for usage tracking.

  45. 48

    dsh-usage-stats

    176Ychris12138/dsh-usage-stats · JavaScript

    dsh-usage-stats is a plugin for the DeepSeek Harness Web GUI (dsh web) that adds a sidebar panel showing provider account balances, subscription quotas, and token-usage analytics across multiple AI providers.

Often tagged together