Jotsu MCP is a general-purpose library for implementing the Model Context Protocol (MCP) and building workflows that leverage MCP tools, resources, and prompts. It provides both a Python library and a CLI you can use to create, run, and test workflows that interact with MCP servers and their tools. With this package installed, you can initialize a workflow, connect to MCP-enabled servers, and execute nodes that call tools on those servers or produce output nodes for debugging and results collection.

The CLI, installed as part of the package (via the [cli] extras), lets you run commands like workflow init and workflow run, enabling rapid prototyping of MCP-driven workflows. The library also supports programmatic usage if you prefer integrating MCP workflow orchestration directly into Python code. The focus is on composing workflows that orchestrate tools from MCP servers and surface the final results in a structured workflow-end node.

You can point your workflow at any MCP-enabled server by adding a server entry with its id, name, and URL, then reference that server's tools in your workflow nodes. This makes it suitable for models that don’t have native MCP support or for orchestrating multiple tools across different servers in a single workflow.

Prerequisites:

• Python 3.8+ (recommended)
• pip (comes with Python)

Installation steps:

1. Optional: create and activate a virtual environment python -m venv venv

   macOS/Linux

   source venv/bin/activate

   Windows

   .\venv\Scripts\activate.bat

2. Install the Jotsu MCP package with CLI support pip install jotsu-mcp[cli]

3. Verify installation jotsu-mcp --version

4. Basic usage to start a workflow (examples, adjust to your setup) jotsu-mcp workflow init jotsu-mcp workflow run ./workflow.json

Optional: if you’re integrating into an application, you can also use the Python API directly by importing the library from jotsu_mcp.

Tips and common considerations:

• If you’re connecting to external MCP servers, ensure URLs are accessible and consider any authentication requirements the target server may impose.
• The extra CLI features are provided via the [cli] extras during installation; omit them if you don’t need the CLI.
• When defining workflow data, you can preload initial data for tools that require context (as shown in the Hello MCP example).
• Workflow end results are surfaced in the final 'result' field of the workflow-end node; workflows can branch, so expected outputs may vary between runs.
• For development, you can install extra dev dependencies as shown in the repository’s development notes, including common tooling for linting and testing.