home / mcp / semiprocess mcp server

SemiProcess MCP Server

Provides universal semiconductor process analyses driven by user data via a streamable MCP API.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "jkwjkw1020-bot-semiprocess_mcp": {
      "url": "https://semiprocess-mcp.vercel.app/mcp"
    }
  }
}

SemiProcess MCP Server provides a universal framework for semiconductor process analysis. It runs statelessly, accepts user-provided data as the basis for 15 configurable analyses, and exposes results through a scalable MCP API with real-time streaming support. It is designed to be deployable publicly and integrates with standard data workflows to help you derive actionable insights from your process knowledge and historical data.

How to use

You connect your MCP client to the server to run analyses and generate reports. You can execute all available tools by sending requests through the MCP API. The API returns results as Markdown, including any warnings, diagnostics, or recommended actions. You access the health endpoint to confirm the service is ready, and you can monitor ongoing analyses via server-sent events (SSE) for live updates.

How to install

Prerequisites: you need Python 3.11 or newer and an environment capable of running FastAPI-based services.

pip install -r requirements.txt

Run the server locally with a configurable port. The port can be changed via PORT in your environment.

export PORT=8000
uvicorn src.server:app --host 0.0.0.0 --port ${PORT}

Additional sections

Configuration essentials include the HTTP MCP endpoint for remote usage and the local stdio command for development. If you run remotely, you will typically connect to the provided HTTP URL; for local development, you start the server with the uvicorn command above.

Security and deployment notes: operate the server behind standard network protections. Enable TLS/HTTPS in production and manage access via your chosen auth strategy. The service is stateless and can be horizontally scaled behind a reverse proxy.

Examples of common deployment patterns include containerized runs (see Docker) and public hosting (see Vercel). The system exposes a health check on GET /health and the MCP interface on GET|POST /mcp, with a service metadata endpoint at GET /.

Troubleshooting tips: if you do not see expected results, verify the incoming data format, check the availability of the HTTP endpoint, and ensure the port is accessible from your client network. For streaming updates, confirm that your client correctly handles server-sent events.

Available tools

analyze_defect

Assess defects using defect_code, defect_description, and process_step data to identify likely causes, generate a checklist, and suggest verification steps.

get_defect_history

Retrieve defect history records to identify trends and recurring issues across lots, tools, or equipment.

suggest_corrective_action

Provide immediate corrective actions for a described problem, including priority steps and escalation guidance.

compare_to_baseline

Compare a current recipe against a baseline recipe to identify deviations and recommended adjustments.

compare_two_recipes

Compare two recipe configurations and determine where tolerances are exceeded.

validate_process_window

Check whether a process window satisfies test conditions and identify critical parameters that require attention.

analyze_metrics

Evaluate key performance indicators against targets over a specified period and report KPI status.

analyze_spc_data

Analyze SPC data with control and specification limits, reporting Cp/Cpk/Pp/Ppk and stability indicators.

predict_defect_risk

Estimate risk levels for defects given process windows and current conditions, incorporating severity and detection metrics.

optimize_recipe_direction

Suggest optimal parameter adjustments to close performance gaps while considering parameter sensitivity.

simulate_parameter_change

Forecast the impact of proposed parameter changes on performance metrics before applying them.

calculate_yield_impact

Predict yield changes due to parameter variations using DOE-inspired logic and confidence intervals.

analyze_equipment_comparison

Rank equipment by performance metrics using weighted comparisons and normalization.

generate_shift_report

Create a comprehensive shift report summarizing production, equipment status, quality, and events.

analyze_trend

Identify trends in time-series data, detect shifts, and summarize predictive implications.