ASAM ODS Jaquel MCP Server

A Model Context Protocol (MCP) server for ASAM ODS with odsbox Jaquel query tools, ODS connection management, and measurement data access.

Overview

• 🔌 Built-in ODS connection management
• 🧰 MCP tools: schema inspection, query validation, direct ODS query execution and measurement data analysis
• 🏗️ Entity hierarchy visualization (AoTest → AoMeasurement)
• 🚀 Validate, explain and execute JAQueL queries for ASAM ODS
• 📦 Bulk timeseries/submatrix data access and script generation
• 📊 Automatic Jupyter notebook generation for measurement comparison
• 📈 Matplotlib visualization code generation
• 📉 Statistical measurement comparison and correlation analysis
• 🔎 Measurement hierarchy exploration and discovery
• 💡 Interactive starting prompts for guided workflows
• 🤖 AI-guided bulk API learning with help_bulk_api tool
• 📝 Comprehensive documentation and test suite

Documentation

• Prompts Guide: See PROMPTS.md for starting prompts documentation
• Tool Reference: See TOOLS_GUIDE.md
• Changelog: See CHANGELOG.md

Quick Start

Installation

Using uvx (Recommended)

The easiest way to use this MCP server is with uvx:

uvx odsbox-jaquel-mcp@latest

This automatically installs and runs the server without managing virtual environments.

Using uv tool (persistent install)

For a persistent installation:

uv tool install odsbox-jaquel-mcp
odsbox-jaquel-mcp

Using project virtual environment

uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
uv pip install odsbox-jaquel-mcp[play]

Note: The [play] extra includes optional data analysis and visualization dependencies (pandas, matplotlib, scipy) for working with Jupyter notebooks and data analysis.

Running the Server

The server runs on stdin/stdout and waits for MCP messages from an MCP client:

# With uvx (auto-installs and runs)
uvx odsbox-jaquel-mcp@latest

# With uv tool (if installed)
odsbox-jaquel-mcp

# With project environment
uv run python -m odsbox_jaquel_mcp

Configuration for MCP Clients

Add to your MCP client configuration (e.g., Claude Desktop, VS Code):

{
  "mcpServers": {
    "ods-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": ["odsbox-jaquel-mcp@latest"]
    }
  }
}

Or with uv tool:

{
  "mcpServers": {
    "ods-mcp": {
      "type": "stdio",
      "command": "odsbox-jaquel-mcp"
    }
  }
}

Environment Variables

See TOOLS_GUIDE.md for the full list of authentication variables and keyring fallback details.

Secrets are looked up in the ods-pilot keyring service first using <service>::<username> records, then in the legacy direct service lookup if no entry is found.

Usage Monitoring

When ODSBOX_STATS_ENABLED=true is set, the server records tool call and resource read statistics to a local SQLite database:

• Location: ~/.local/share/odsbox-jaquel-mcp/odsbox-jaquel-mcp-stats.db (Linux/macOS) or %APPDATA%\odsbox-jaquel-mcp\odsbox-jaquel-mcp-stats.db (Windows), with fallback to the system temp directory.
• Tracked per tool: call count, error count, total execution time (ms), last called timestamp.
• Tracked per resource: read count, error count, total execution time (ms), last read timestamp.
• Cross-process safe: uses SQLite WAL mode, so multiple concurrent MCP sessions can write safely.

You can query the stats database directly:

sqlite3 ~/.local/share/odsbox-jaquel-mcp/odsbox-jaquel-mcp-stats.db \
  "SELECT name, calls, errors, total_ms FROM tool_stats ORDER BY calls DESC"

Example MCP client configuration with monitoring enabled:

{
  "mcpServers": {
    "ods-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": ["odsbox-jaquel-mcp@latest"],
      "env": {
        "ODSBOX_STATS_ENABLED": "true",
        "FASTMCP_LOG_LEVEL": "WARNING"
      }
    }
  }
}

Development

Setup

git clone https://github.com/totonga/odsbox-jaquel-mcp.git
cd odsbox-jaquel-mcp
uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
uv sync --all-extras --group dev

Common Tasks

# Run server locally
uv run python -m odsbox_jaquel_mcp

# Run tests
uv run pytest tests/
# or
python run_tests.py

# Code formatting and linting
ruff check .
ruff format .

# Build package
uv build

# Test with MCP Inspector
npx @modelcontextprotocol/inspector uvx odsbox-jaquel-mcp@latest

Contributing

Pull requests and issues are welcome! Please:

• Follow PEP8 and use type hints
• Add/maintain tests for new features
• Update documentation as needed

License

This project is licensed under the Apache License 2.0. See LICENSE.

Links

• ASAM ODS
• MCP Protocol
• odsbox

Features

Core MCP Tools

Connection Management

• ods_connect - Establish ODS connection
• ods_connect_using_env - Establish ODS connection using environment variables
• ods_disconnect - Close ODS connection
• ods_get_connection_info - Get connection status

Schema Inspection

• schema_get_entity - Get all fields for entity
• schema_list_entities - List all entities with relationships
• schema_test_to_measurement_hierarchy - Get ASAM ODS test hierarchy structure

Query Building & Validation

• query_validate - Check query syntax and structure
• query_describe - Get plain English explanation
• query_execute - Execute query on ODS server

Timeseries/Submatrix Data Access

• data_get_quantities - List measurement quantities for submatrix
• data_read_submatrix - Read timeseries data from submatrix
• data_generate_fetcher_script - Generate Python scripts for data fetching

Pattern & Example Library

• query_generate_skeleton - Generate query skeleton (basic query) for entity
• query_get_pattern - Get template for common patterns
• query_list_patterns - List available patterns
• query_get_operator_docs - Learn about operators

Starting Prompts

Discover and use the server's capabilities through interactive guided prompts:

• ODS Server Connection - Set up and manage connections
• Validate a Jaquel Query - Learn query validation
• Explore Query Patterns - Find common query templates
• Bulk Data Access - Master the 3-step Bulk API workflow
• Measurement Analysis - Compare measurements and visualize data

See PROMPTS.md for complete details on all starting prompts.

Error Handling

Common Errors and Solutions

Not connected

{
  "error": "Model not loaded",
  "hint": "Connect to ODS server using 'ods_connect' tool first"
}

Solution: Call ods_connect first

Invalid entity

{
  "error": "Entity not found: InvalidEntity",
  "available_entities": ["AoUnit", "AoMeasurement", ...]
}

Solution: Use valid entity from available_entities

Invalid field

{
  "valid": false,
  "issues": ["Field 'invalid_field' not found"],
  "suggestions": ["id", "name", "description"]
}

Solution: Use one of the suggested fields

Connection failed

{
  "success": false,
  "error": "Connection refused",
  "error_type": "ConnectionError"
}

Solution: Check URL, server availability, firewall

Troubleshooting

Issue: Tools not discovered

• Ensure mcp>=0.1.0 is installed
• Check ToolsCapability is set in ServerCapabilities
• Restart MCP client