# token-compressor

Compress prompts 40-60% using local LLM + embedding validation. Preserves all conditionals.

- Weekly downloads: 29
- GitHub stars: 12
- Health: at-risk
- Repository: https://github.com/base76-research-lab/token-compressor
- Registry names: io.github.base76-research-lab/token-compressor
- Categories: Search & knowledge, Dev tools & GitHub, AI models & media
- Snapshot: 2026-10-14

## Install

```sh
uvx token-compressor-mcp
```

```json
{
  "mcpServers": {
    "token-compressor": {
      "command": "uvx",
      "args": [
        "token-compressor-mcp"
      ]
    }
  }
}
```

Source: https://mcpmorning.com/mcp/base76-research-lab-token-compressor/