Skip to main content

Arcanna MCP Server

An MCP server implementation for interacting with Arcanna's AI-powered decision-making system. This server provides tools for managing jobs, sending events for analysis, and handling feedback for continuous learning.

Configuration

Setting up Arcanna Credentials

  1. Generate a Management API key
  2. Set your MCP client to connect to the desired Arcanna instance.

Prerequisites

  1. Make sure you have docker up and running.
  2. Make sure you have an AI job in arcanna that has an External REST API as input. The API KEY you used for the External REST API will be the one you set up in arcanna's mcp server environment.

Usage with Claude Desktop

Add this config to your claude_desktop_config.json:

{
"mcpServers": {
"arcanna-mcp-stdio": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"ARCANNA_MANAGEMENT_API_KEY",
"-e",
"ARCANNA_HOST",
"arcanna/arcanna-mcp-server"
],
"env": {
"ARCANNA_MANAGEMENT_API_KEY": "<ARCANNA_MANAGEMENT_API_KEY>",
"ARCANNA_HOST": "<YOUR_ARCANNA_HOST_HERE>"
}
}
}
}

Features

  • Resource Management: Create, update and retrieve Arcanna resources (jobs, integrations)
  • Python Coding: Code generation, execution and save the code block as Arcanna integration
  • Query Arcanna events: Query events processed by Arcanna
  • Job Management: Create, retrieve, start, stop, and train jobs
  • Feedback System: Provide feedback on decisions to improve model accuracy
  • Health Monitoring: Check server and API key status

Tools

Query Arcanna events

  • query_arcanna_events

    • Used to get events processed by Arcanna, multiple filters can be provided
  • get_filter_fields

    • used as a helper tool (retrieve Arcanna possible fields to apply filters on)

Resource Management

  • upsert_resources

    • Create/update Arcanna resources
  • get_resources

    • Retrieve Arcanna resources (jobs/integrations)
  • delete_resources

    • Delete Arcanna resources
  • integration_parameters_schema

    • used in this context as a helper tool

Python Coding

  • generate_code_agent

    • Used to generate code
  • execute_code

    • Used to execute the generated code
  • save_code

    • Use to save the code block in Arcanna pipeline as an integration

Job Management

  • start_job

    • Begin event ingestion for a job
  • stop_job

    • Stop event ingestion for a job
  • train_job

    • Train the job's AI model using the provided feedback

Feedback System

  • add_feedback_to_event
    • Provide feedback on AI decisions for model improvement

System Health

  • health_check
    • Verify server status and API key validity
    • Returns API key authorization status

Usage example