AI Tools Setup¶
Tracker includes an MCP (Model Context Protocol) server that gives AI assistants read-only query access to the workflow database.
MCP Server URL¶
https://mcp.tracker.292967571998.natera.io/mcp/
Available Tools¶
| Tool | Description |
|---|---|
schema |
Returns the column-level database schema. Call this before writing SQL — it exposes the tables, columns, and data types your queries can use. |
query |
Read-only SQL queries against the Tracker database (runs, tasks, processes, outputs, metadata). |
cost_and_performance |
Cost and performance analysis for a run (same data as the Cost & Performance tab). |
qc_report_summary |
Path to the MultiQC llms-full.txt summary for a run, when available. |
ai_analysis_result |
The Markdown body of the auto-generated AI Analysis for a run -- QC summary for succeeded runs, failure investigation for failed runs. |
url |
Builds a Tracker URL for a run (optionally a tab, task, attempt, or process). Use to generate deep links in your replies. |
Setup¶
Claude Code CLI¶
claude mcp add --transport http workflow-tracker https://mcp.tracker.292967571998.natera.io/mcp/
JSON Configuration¶
Add to ~/.claude/settings.json (global) or .mcp.json (project-level):
{
"mcpServers": {
"workflow-tracker": {
"type": "http",
"url": "https://mcp.tracker.292967571998.natera.io/mcp/"
}
}
}
For other MCP-compatible tools (Cursor, Windsurf, etc.), use transport type HTTP with the URL above.
Example Queries¶
- "What's the status of run X? Give me process-level summaries."
- "Compare runtimes and resource usage of runs X and Y."
- "Show me a histogram of BWAMEM runtimes for all
daq-*runs." - "Which runs failed in the last 24 hours? What were the error messages?"
- "What is the average memory usage of FASTQC across recent DAQ runs?"