Token efficiency: tools for developers

Tools that reduce token costs, compress context or cache LLM requests. An entry can appear under several use cases when its topics or description provide several explicit signals.

Use case

Activity

Sort by

98 entries in this view.

Tool ranking

Ranked by measured growth, normalized across sources. Raw values keep their source-specific window, shown under each tool; a change requires at least two snapshots from the last 7 days.

  1. 1

    Skill
    Active

    Resume Claude Code work after rate/usage/context limits without replaying the prior transcript. Auto-saves at 90%/95% usage. Plugin-installable, 10 languages.

    PowerShellLicense: MIT
    • checkpoint
    • claude-code
    • claude-code-plugin
    • claude-code-skills
    • context

    githubmeasured growthOpen source ↗

    Install /plugin marketplace add sofumel/claude-handoff-revive

    6GitHub starsstable
  2. 2

    Skill
    Active

    Token-efficient Claude Code workspace with parallel agents and persistent memory. Research → Plan → Implement → Validate workflow.

    JavaScript
    • agent-memory
    • agentic-coding
    • ai-agents
    • ai-coding
    • claude-code-commands

    githubmeasured growthOpen source ↗

    Install /plugin marketplace add coalesce-labs/catalyst

    20GitHub starsstable
  3. 3

    MCP
    Active

    Lossless context compression: 2-8x fewer tokens, byte-exact recovery, search inside payloads

    PythonLicense: MIT

      mcpmeasured growthOpen source ↗

      Install claude mcp add densely -- uvx densely

      6GitHub starsstable
    • 4

      Skill
      Active

      Token-compression skill. An adaptation of caveman — short common words, trust context, say just enough, be laconic.

      PythonLicense: MIT
      • ai-assistant
      • anthropic
      • brevity
      • claude
      • claude-code

      githubmeasured growthOpen source ↗

      Install /plugin marketplace add GabrielBarberini/laconic

      19GitHub starsstable
    • 5

      Skill
      Active

      Local-first memory and prompt-cache layer for Claude Code

      TypeScriptLicense: MIT
      • agentic-ai
      • agentic-coding
      • ai
      • ai-agents
      • anthropic

      githubmeasured growthOpen source ↗

      Install /plugin marketplace add riz007/somtum

      7GitHub starsstable
    • 6

      MCP
      Active

      A MCP that turns PDFs into structured text, tables, figures, metadata, annotations, and page-level evidence for selective retrieval by MCP-compatible AI agents. Improves the information quality AI agents receive while also reducing token…

      JavaScriptLicense: AGPL-3.0
      • ai
      • ai-tools
      • claude
      • claude-code
      • claude-code-plugin

      githubmeasured growthOpen source ↗

      17GitHub starsstable
    • 7

      Skill
      Active

      A production-ready Claude Code setup. Global CLAUDE.md, coding rules, per-project memory, and token optimization.

      ShellLicense: MIT
      • ai-agents
      • claude
      • claude-code
      • claude-code-plugin
      • context-engineering

      githubmeasured growthOpen source ↗

      Install /plugin marketplace add janmaaarc/basecamp

      7GitHub starsstable
    • 8

      Skill
      Active

      Claude Code skill — delegate coding tasks to Opencode, supervise via git diff & save 50-90% Claude token usage

      Python
      • claude
      • claude-skills
      • delegation
      • opencode
      • orchestration

      githubmeasured growthOpen source ↗

      Install git clone https://github.com/pcx-wave/opencode-skill ~/.claude/skills/opencode-skill

      17GitHub stars+1 (+6.2 %)
    • 9

      Agent
      Active

      Control plane for AI coding agents: route tasks, reduce token spend, run multi-agent workflows, fallback executors, and track cost per task.

      PythonLicense: Apache-2.0
      • a2a
      • ag-ui
      • agentic-ai
      • ai-agents
      • ai-coding

      githubmeasured growthOpen source ↗

      16GitHub starsstable
    • 10

      MCP
      Active

      Review and compact prompts, docs, and AI-agent skills to save tokens, preserving structure.

      ↓ 27 downloads/weekPythonLicense: MIT

        mcpmeasured growthOpen source ↗

        Install claude mcp add compactprompt -- uvx compactprompt

        0GitHub starsstable
      • 11

        Skill
        Active

        mcp-recall compresses MCP tool outputs (94 KB → 3.5 KB · 96%) and stores full results in SQLite for retrieval — up to 30x more tool calls per session for heavy MCP workloads.

        TypeScriptLicense: MIT
        • bun
        • claude
        • claude-code
        • claude-code-plugin
        • context-window

        githubmeasured growthOpen source ↗

        Install /plugin marketplace add sakebomb/mcp-recall

        8GitHub starsstable
      • 12

        MCP
        Dormant

        DAG-based Lossless Context Management for Claude Code. Every message preserved, summaries cascade as a DAG, patterns extracted via Lossless Dream — full recall and reflection across sessions.

        PythonLicense: MIT
        • claude
        • claude-code
        • claude-code-plugin
        • coding-agent
        • context-management

        githubmeasured growthOpen source ↗

        15GitHub starsstable
      • 13

        Skill
        Active

        Compact Claude Code statusline — zero-dependency, crash-free on Windows

        TypeScriptLicense: MIT
        • claude
        • claude-code
        • claude-code-plugin
        • context-window
        • deepseek

        githubmeasured growthOpen source ↗

        Install /plugin marketplace add WaterTian/cc-hud

        8GitHub starsstable
      • 14

        Skill
        Active

        Staged context pruning for OpenCode, Oh My Pi, pi, and Claude Code.

        TypeScriptLicense: AGPL-3.0
        • ai
        • ai-agents
        • claude-code
        • claude-code-plugin
        • coding-agent

        githubmeasured growthOpen source ↗

        Install /plugin marketplace add AshishKumar4/better-compact

        12GitHub stars+1 (+9.1 %)
      • 15

        MCP
        Active

        Persistent memory + security layer for Claude Code (MCP plugin). Agents remember your project across sessions (~87% fewer context tokens), every tool call lands in a tamper-evident HMAC audit chain, and skills are AST-scanned +…

        TypeScriptLicense: MIT
        • agent-security
        • ai-agent-memory
        • ai-agents
        • audit-trail
        • claude-code

        githubmeasured growthOpen source ↗

        8GitHub starsstable
      • 16

        Skill
        Dormant

        Say more with less. Claude Code skill/plugin that cuts ~60-70% of output tokens while keeping natural, readable English.

        ShellLicense: MIT
        • ai-coding
        • claude-code
        • claude-code-plugin
        • claude-code-plugins
        • claude-code-skill

        githubmeasured growthOpen source ↗

        Install /plugin marketplace add o4f6bgpac3/concise

        10GitHub stars+2 (+25.0 %)
      • 17

        CLI
        Dormant

        Optimize Claude Code token usage — 22 skills, 8 hooks, 6 rules. One npm install. Zero manual steps. Cut costs by 67%.

        ShellLicense: MIT
        • ai
        • ai-coding
        • anthropic
        • claude
        • claude-code

        githubmeasured growthOpen source ↗

        9GitHub starsstable
      • 18

        Skill
        Active

        对标 claude-hud 核心功能，控制在 7 行以内更轻量。内置 13 个国产大模型平台用量查询 (智谱 GLM / MiniMax / DeepSeek / Kimi / StepFun / SiliconFlow 等)，第三方代理用户也能实时看到额度消耗。

        TypeScriptLicense: MIT
        • anthropic
        • claude
        • claude-code
        • claude-code-plugin
        • cli

        githubmeasured growthOpen source ↗

        Install /plugin marketplace add zander-zyx/claude-mini-hud

        9GitHub starsstable
      • 19

        MCP
        Active

        Context GC for LLM agents: offload large tool outputs and recall them to save tokens.

        PythonLicense: Unlicense

          mcpmeasured growthOpen source ↗

          Install claude mcp add lethe -- uvx lethe-llm-context

          0GitHub starsstable
        • 20

          Skill
          Dormant

          Drop-in Cursor rules to cut AI response tokens (no filler/hedging, preserve code + technical strings). Includes local, reproducible benchmarks.

          JavaScriptLicense: MIT
          • benchmarking
          • bonsai
          • cursor
          • cursor-rules
          • developer-tools

          githubmeasured growthOpen source ↗

          Install git clone https://github.com/marcomondelli/bonsai ~/.claude/skills/bonsai

          0GitHub starsstable
        • 21

          MCP
          Active

          Statically audits MCP tool surfaces for token cost, schema quality, and design issues.

          TypeScript

            mcpmeasured growthOpen source ↗

            Remote server, to be declared in the MCP configuration https://mcplint-web.vercel.app/api/mcp

            0GitHub starsstable
          • 22

            MCP
            Active

            Verifies token/cost savings claimed by AI context-reduction proxies via MCP tools.

            TypeScriptLicense: Apache-2.0

              mcpmeasured growthOpen source ↗

              Install claude mcp add tokentrust -- uvx tokentrust-cli

              0GitHub starsstable
            • 23

              MCP
              Active

              Reversible context compression for AI agents: cut token usage up to 94%, originals retrievable

              ↓ 13 downloads/weekPythonLicense: Apache-2.0

                mcpmeasured growthOpen source ↗

                Install claude mcp add token_optimizer -- uvx slimctx

                1GitHub starsstable
              • 24

                MCP
                Active

                Cut AI agent token costs 5-15x — routes only relevant code symbols instead of full files.

                ↓ 55 downloads/weekPythonLicense: MIT

                  mcpmeasured growthOpen source ↗

                  Install claude mcp add agent-booster -- uvx agent-booster

                  1GitHub starsstable

                Listed without a public metric

                Mixed ranking: measured growth takes priority; entries without two snapshots are estimated. Raw values keep their source-specific window, shown under each tool; a change requires at least two snapshots from the last 7 days.

                1. 1

                  MCP
                  UnknownLast commit: unavailable

                  Cut LLM token costs: count tokens, estimate cost, slim prompts, and pick the cheapest capable model.

                  mcpmetric not publishedOpen source ↗

                  Install claude mcp add mcp-token-optimizer -- npx mcp-token-optimizer

                  —No public metric provided