by ruco

Few tokens.
One server.

MCP server that cuts 65% of tokens by compressing markdown files and activating caveman speak. No file distribution. No sync. No API key. Register once — works everywhere.

65% tokens purged

0 files to sync

∞ projects covered

5 agent clients

❯ pip install caveman-mcp

or run without installing: uvx caveman-mcp

01 // Why MCP

The MCP Advantage

Caveman prompts used to require a file in every project. MCP changes everything.

🔌

Register Once

Add caveman-mcp to your client config once. It activates across all your projects, forever. No per-repo setup.

🚫

No File Sync

The original caveman needed a prompt file in every repo. No more copy-paste. No stale prompts. No drift.

🔑

No API Key

Zero credentials. Zero auth friction. Runs as a local stdio server — your data stays on your machine.

🛠️

Tools Included

MCP exposes compress tools directly to the agent. compress_prepare → compress_write → compress_restore. The agent does the work.

🌐

Any Client

Works with Claude Code, Cursor, Windsurf, Cline, and any MCP-compatible host. One install, five agents.

⚡

Auto-Compress Hook

Wire up a Claude Code PostToolUse hook and every markdown file gets compressed automatically on read. Invisible. Zero effort.

02 // Compression Matrix

The Lithic Payload Diff

Payload: React Re-render Diagnosis

Without caveman-mcp 69 tokens

"The reason your React component is re-rendering is likely because you're creating a new object reference on each render cycle. When you pass an inline object as a prop, React's shallow comparison sees it as a different object every time, which triggers a re-render. I'd recommend using useMemo to memoize the object."

With caveman-mcp 19 tokens

"New object ref each render. Inline object prop = new ref = re-render. Wrap in `useMemo`."

Same fix. 75% fewer tokens.

Explain React re-render bug

87%

Fix auth token expiry

83%

Debug PostgreSQL race condition

81%

Explain git rebase vs merge

58%

Review PR for security

41%

Average 65% reduction. Reasoning tokens unaffected. Output payload only.

03 // Intensity Levels

Compression Spectrum

Six modes from terse to ancient. Pick your poison.

Mode	Effect
lite	Drop filler. Keep full sentences and articles.
full	Default — drop articles, fragments OK, short synonyms.
ultra	Abbreviate (DB/auth/req/res/fn), strip conjunctions, X→Y causality.
wenyan-lite	Semi-classical Chinese register.
wenyan-full	Full 文言文 — 80–90% character reduction.
wenyan-ultra	Extreme. Ancient scholar feel.

Available Prompts

/caveman

Activate caveman compression

/caveman-commit

Terse commit message style

/caveman-review

One-line code review comments

/caveman-help

Quick-reference card

04 // Initialize Protocol

Connect Your Client

Claude Code Claude Desktop Cursor Windsurf Cline

Claude Code — global config (recommended):

// ~/.claude/settings.json
{
  "mcpServers": {
    "caveman-mcp": {
      "command": "uvx",
      "args": ["caveman-mcp"]
    }
  }
}

Or add via CLI:

claude mcp add caveman-mcp uvx -- caveman-mcp

Activate in any session:

❯ /caveman

Me ready.