prosuite-mcp

MCP server that exposes Dira ProSuite quality verification to AI assistants (Claude, etc.).

Prerequisites

A running ProSuite Quality Verification Server reachable from the host where this server runs.

Configuration

Usage

Windows users: see docs/windows-setup.md for a step-by-step guide including uv and Claude Code installation.

Both options below assume you create a project directory first:

mkdir mytest
cd mytest
uv init --python 3.12
uv add prosuite-mcp

Claude Code CLI

Register the server from inside mytest, then start Claude:

claude mcp add prosuite \
  -e PROSUITE_HOST=localhost \
  -e PROSUITE_PORT=5151 \
  -- uv run prosuite-mcp

claude

The -- uv run prosuite-mcp tells Claude Code to start the MCP server via uv run in the current project, so prosuite-mcp is resolved from the local .venv. Run claude from the same mytest directory each time.

Copilot CLI

Register the server from inside mytest, then start Copilot:

copilot mcp add prosuite \
  -e PROSUITE_HOST=localhost \
  -e PROSUITE_PORT=5151 \
  -- uv run prosuite-mcp

Tools

load_spec <path> — Loads a .qa.xml spec file. Call this at the start of a session with the path to your spec (local drive, OneDrive, network share). Replaces any previously loaded spec.

search_spec <query> [max_results] — Searches the loaded .qa.xml spec for conditions matching a natural-language query (English, German, French, Italian). Returns up to max_results (default 20) matching conditions with pre-filled condition_request blocks ready to pass directly to run_verification, plus the required_datasets list. Requires a spec to be loaded first via load_spec.

list_conditions [search] — Lists available quality conditions. Pass a keyword to filter by name or description.

describe_condition <name> — Shows the full docstring and parameter list for a condition, including which parameters expect dataset names vs. primitive values.

run_verification — Runs an ad-hoc quality verification against a workspace. Key parameters:

Returns a summary with status, total_errors, and a per-condition breakdown.

Example

Once connected, you talk to Claude in plain language:

Check road connectivity in C:/data/tlm.sde.

With a spec loaded, Claude calls search_spec to find the relevant pre-configured conditions from the .qa.xml file, then calls run_verification with the pre-filled parameters and returns a summary of errors per condition.

Without a spec, Claude uses list_conditions and describe_condition to find and configure conditions from scratch.

Development

uv sync --dev
uv run pytest
uv run ruff check src
uv run pyright src

License

MIT — see LICENSE.