PyTorch HUD MCP Server
by izaitsevfb
hybrid server
The server is able to function both locally and remotely, depending on the configuration or use case.
Integrations
Provides ClickHouse query integration for analytics of PyTorch CI/CD data
Provides access to PyTorch CI/CD data, job logs, and analytics
PyTorch HUD API with MCP Support
A Python library and MCP server for interacting with the PyTorch HUD API, providing access to CI/CD data, job logs, and analytics.
Overview
This project provides tools for PyTorch CI/CD analytics including:
- Data access for workflows, jobs, and test runs
- Efficient log analysis for large CI logs
- ClickHouse query integration for analytics
- Resource utilization metrics
Usage (for humans)
Development
Key Features
Data Access
get_commit_summary
: Basic commit info without jobsget_job_summary
: Aggregated job status countsget_filtered_jobs
: Jobs with filtering by status/workflow/nameget_failure_details
: Failed jobs with detailed failure infoget_recent_commit_status
: Status for recent commits with job statistics
Log Analysis
download_log_to_file
: Download logs to local storageextract_log_patterns
: Find errors, warnings, etc.extract_test_results
: Parse test execution resultsfilter_log_sections
: Extract specific log sectionssearch_logs
: Search across multiple logs
Development
Documentation
- CLAUDE.md: Detailed usage, code style, and implementation notes
- mcp-guide.md: General MCP protocol information
License
MIT
This server cannot be installed
Provides access to PyTorch CI/CD analytics data including workflows, jobs, test runs, and log analysis through an MCP interface.