Decompose

Provides an MCP server that deconstructs text into structured semantic units for faster, reliable downstream reasoning.

python

1

GitHub Stars

python

Language

4 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "echology-io-decompose": {
      "command": "uvx",
      "args": [
        "decompose-mcp",
        "--serve"
      ]
    }
  }
}

Decompose MCP Server provides a structured, fast preprocessing step for your agents by turning text into classified semantic units. You can run it locally and expose an MCP endpoint that your client can query to obtain decomposed data, enabling faster, more reliable downstream reasoning without relying on large language models.

How to use

To use the Decompose MCP Server, run one of the provided local MCP server styles and point your MCP client to the server as you would with any standard MCP service. You don’t need to embed complex preprocessing in your client; instead, you receive a structured, ready-to-use decomposition that your agent can reason over.

How to install

Prerequisites you need on your machine before installing include Python 3 and a working Python package manager. You can also use a containerized or alternative runtime if you prefer, but the following steps assume a Python environment.

pip install decompose-mcp
python3 -m pip show decompose-mcp  # verify installation
# Optional: upgrade to the latest patch release
python3 -m pip install --upgrade decompose-mcp

# If you want to run the server directly with Python, you can follow the OpenClaw approach below.

OpenClaw style local server (Python)

Run the server in a local, stdio-based mode using Python. This starts an MCP server that your client can connect to via the standard MCP protocol on its standard input/output streams.

{
  "mcpServers": {
    "decompose": {
      "command": "python3",
      "args": ["-m", "decompose", "--serve"]
    }
  }
}

UVX style local server

If you prefer a lightweight launcher, you can run the server with the UVX wrapper. This starts the MCP server in serve mode so your agent can send requests and receive decomposed outputs.

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

Available tools

decompose_text

Decompose a plain text snippet into a list of structured units with fields like authority, risk, type, irreducible, attention, and entities.

decompose_url

Fetch a URL and decompose its content into structured units, returning the same type of structured output for downstream use.