A Model Context Protocol (MCP) server that provides tools for connecting LLMs to Itential Platform. Enable AI assistants to manage network automations, orchestrate workflows, and monitor platform health.
- Multiple Transport Methods: Choose between stdio (default) or SSE transport for MCP server
- Dynamic Tool Loading: Automatically discovers and registers tools without modifying core code
- Flexible Authentication: Supports both basic authentication and OAuth for Itential Platform
- Configurable: Set options via command line parameters or environment variables
- Containerized: Run as a Docker container with configurable environment
- Extensible: Easy to add new tools without deep knowledge of the code base
- Python 3.10 or higher
- Access to an Itential Platform Instance
- For development -
uvandmake
The itential-mcp application can be installed using either PyPI or it can be
run directly from source.
To install it from PyPI, simply use pip:
pip install itential-mcpThe repository can also be clone the repository to your local environment to
work with the MCP server. The project uses uv and make so both tools
would need to be installed and available in your environment.
The following commands can be used to get started.
git clone https://github.com/itential/itential-mcp
cd itential-mcp
make buildBuild and run as a container:
# Build the container image
make container
# Run the container with environment variables
docker run -p 8000:8000 \
--env ITENTIAL_MCP_TRANSPORT=sse \
--env ITENTIAL_MCP_HOST=0.0.0.0 \
--env ITENTIAL_MCP_PORT=8000 \
--env ITENTIAL_MCP_PLATFORM_HOST=URL \
--env ITENTIAL_MCP_PLATFORM_CLIENT_ID=CLIENT_ID \
--env ITENTIAL_MCP_PLATFORM_CLIENT_SECRET=CLIENT_SECRET \
itential-mcp:develStart the MCP server with default settings (stdio transport):
itential-mcp --transport --host 0.0.0.0 --port 8000Start with SSE transport:
itential-mcp --transport sse --host 0.0.0.0 --port 8000| Option | Description | Default |
|---|---|---|
--transport |
Transport protocol (stdio or sse) | stdio |
--host |
Host address to listen on | localhost |
--port |
Port to listen on | 8000 |
--log-level |
Log level (DEBUG, INFO, WARNING, ERROR, CRITICAL) | INFO |
| Option | Description | Default |
|---|---|---|
--platform-host |
Itential Platform hostname | localhost |
--platform-port |
Platform port (0 = auto-detect) | 0 |
--platform-disable-tls |
Disable TLS for platform connection | false |
--platform-disable-verify |
Disable certificate verification | false |
--platform-timeout |
Connection timeout | 30 |
--platform-user |
Username for authentication | admin |
--platform-password |
Password for authentication | admin |
--platform-client-id |
OAuth client ID | none |
--platform-client-secret |
OAuth client secret | none |
All command line options can also be set using environment variables prefixed with ITENTIAL_MCP_. For example:
export ITENTIAL_MCP_TRANSPORT=sse
export ITENTIAL_MCP_PLATFORM_HOST=platform.example.com
itential-mcp # Will use the environment variablesThe MCP server provides the following tools for interaction with Itential Platform:
get_health: Retrieve platform health status including memory, CPU usage, and service status
get_job_metrics: Get aggregated workflow job metricsget_task_metrics: Get aggregated workflow task metrics
get_workflows: List available workflowsstart_workflow: Start a workflow with provided variables and permissions
get_jobs: List all jobs (with optional filtering)describe_job: Get details about a specific job (based on job_id)
get_command_templates: List all configured command templatesdescribe_command_template: Get details about a specific command templaterun_command_template: Run a command template against a set of devices
get_devices: Get a list of devices from Itential Platformrun_command: Run a command against one or more devices
Note: See here for information about how to properly expose workflows to the MCP server
Adding a new tool is simple:
- Create a new Python file in the
src/itential_mcp/tools/directory or add a function to an existing file - Define an async function with a
Contextparameter annotation:
from fastmcp import Context
async def my_new_tool(ctx: Context) -> dict:
"""
Description of what the tool does
Args:
ctx (Context): The FastMCP Context object
Returns:
dict: The response data
Raises:
None
"""
# Get the platform client
client = ctx.request_context.lifespan_context.get("client")
# Make API requests
res = await client.get("/your/api/path")
# Return JSON-serializable results
return res.json()Tools are automatically discovered and registered when the server starts.
Run the test suite with:
make testFor test coverage information:
make coverageContributions are welcome! Please read our Code of Conduct before contributing.
- Fork the repository
- Create a feature branch:
git checkout -b feature/my-feature - Commit your changes:
git commit -am 'Add new feature' - Push to the branch:
git push origin feature/my-feature - Submit a pull request
Before submitting:
- Run
make premergeto ensure tests pass and code style is correct - Add documentation for new features
- Add tests for new functionality
This project is licensed under the GNU General Public License v3.0 - see the LICENSE file for details.
Copyright (c) 2025 Itential, Inc