synthelion 1.0.1

MCP plugin + Python library for LLM token compression. 50+ languages, zero ML models. Port of Caveman (C#).

Synthelion — Token Compressor for Claude Code

Synthelion is a plugin for Claude Code that compresses your prompts before sending them to AI models — saving tokens, reducing costs, and speeding up responses. Works across 50+ languages with no AI model required.

"Why use many tokens when few tokens do trick?" — A caveman (and your wallet).

---

Install in 3 steps

Step 1 — Install the package

pip install synthelion

Step 2 — Add to Claude Code

Open your Claude Code settings file and add the synthelion block inside mcpServers:

macOS / Linux → ~/.claude/settings.json
Windows → %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "synthelion": {
      "command": "synthelion-mcp"
    }
  }
}

Step 3 — Restart Claude Code

Close and reopen Claude Code. Synthelion is ready.

---

No pip? Use uvx instead (zero install)

If you have uv installed, skip pip install entirely:

{
  "mcpServers": {
    "synthelion": {
      "command": "uvx",
      "args": ["synthelion-mcp"]
    }
  }
}

---

What you can ask Claude

Once installed, just talk to Claude naturally:

"Compress this text to save tokens"

"What language is this text written in?"

"Summarize this article in 3 sentences"

"Compress this JSON / HTML / git diff / log file"

"Compress these 10 prompts all at once"

Synthelion automatically picks the best compression strategy for each content type.

---

Tools

Tool	What it does
compress	Removes stop words and lemmatizes content words. Up to 70% token reduction.
detect_language	Identifies the language of any text (returns ISO 639-3 code).
route_content	Auto-detects JSON, HTML, git diff, logs, code or prose and compresses accordingly.
summarize	Extractive summarization — keeps the most important sentences.
compress_batch	Compresses a list of texts in a single call.

---

Compression levels

Level	What it removes	Typical savings
light	Stop words only (articles, prepositions…)	25–35%
semantic	Stop words + reduces words to their base form	30–69%
aggressive	Everything above + generic verbs and adjectives	35–70%

Default is semantic. Ask Claude to use a different level if needed:

"Compress this aggressively to save as many tokens as possible"

---

Troubleshooting

synthelion-mcp: command not found

Python's Scripts folder is not in your PATH. Use the full path instead:

{
  "mcpServers": {
    "synthelion": {
      "command": "python",
      "args": ["-m", "synthelion.plugins.mcp_server"]
    }
  }
}

Or just use the uvx option above — it always works without PATH issues.

---

Supported languages (50+)

Afrikaans, Arabic, Basque, Belarusian, Bengali, Bulgarian, Catalan, Chinese, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, Galician, German, Greek, Hebrew, Hindi, Hungarian, Armenian, Icelandic, Indonesian, Irish, Italian, Japanese, Kannada, Kazakh, Korean, Latin, Latvian, Lithuanian, Macedonian, Malay, Marathi, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Serbian, Slovak, Slovenian, Spanish, Swedish, Tamil, Telugu, Thai, Turkish, Ukrainian, Urdu, Vietnamese.

Language is detected automatically — no configuration needed.

---

PyPI · Source

• PyPI: https://pypi.org/project/synthelion/
• Source: https://github.com/francescopaolopassaro/synthelion
• Original C# project (Caveman): https://github.com/francescopaolopassaro/caveman

© 2026 Passaro Francesco Paolo — Digitalsolutions.it