Back to Browse

Devray Datalab MCP Server

Developer ToolsModerate7.0MCP RegistryLocal
Free

Server data from the Official MCP Registry

Local-first data science and machine learning MCP server

About

Local-first data science and machine learning MCP server

Security Report

7.0
Moderate7.0Low Risk

Valid MCP server (2 strong, 1 medium validity signals). 6 known CVEs in dependencies Imported from the Official MCP Registry. 1 finding(s) downgraded by scanner intelligence.

6 files analyzed · 7 issues found

Security scores are indicators to help you make informed decisions, not guarantees. Always review permissions before connecting any MCP server.

Permissions Required

This plugin requests these system permissions. Most are normal for its category.

env_vars

Check that this permission is expected for this type of plugin.

file_system

Check that this permission is expected for this type of plugin.

How to Install

Add this to your MCP configuration file:

{
  "mcpServers": {
    "devray-datalab-mcp": {
      "command": "uvx",
      "args": [
        "devray-datalab-mcp==0.3.1"
      ]
    }
  }
}

Some servers take additional arguments. See the repository for details.

Documentation

View on GitHub

From the project's GitHub README.

DevRay DataLab MCP

Local-first Data Science & Machine Learning for AI Assistants

PyPI version Python 3.11+ License: MIT CI MCP Registry DevRay DataLab MCP MCP server – quality and maintenance score on Glama Transport: stdio

DevRay DataLab MCP is a local-first Model Context Protocol (MCP) server that lets AI assistants such as OpenAI Codex, Cursor, Claude Desktop, Cline, Kilo Code, and others perform real data science and machine learning on your own computer — no cloud services, no API keys, no paid subscriptions.

Everything runs locally via stdio: dataset loading, exploratory data analysis, preprocessing, classification/regression training, model comparison, export, and prediction. Your datasets never leave your machine.

Installation

Method A — uvx (simplest, no virtualenv)

uvx devray-datalab-mcp

Requires uv. Install uv if needed:

# Windows PowerShell
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

Method B — pip in a virtual environment

python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install devray-datalab-mcp
devray-datalab-mcp

Linux / macOS:

python3 -m venv .venv
source .venv/bin/activate
pip install devray-datalab-mcp
devray-datalab-mcp

Method C — from GitHub source

pip install git+https://github.com/rayklanderman/devray-datalab-mcp.git
devray-datalab-mcp

A stdio MCP server waits for input from an MCP client. Launching devray-datalab-mcp directly without a connected client will appear to do nothing — this is expected. Connect it through an MCP-compatible client (see client configuration below).

Configure where datasets and model artifacts live with environment variables:

VariableDefaultMeaning
DEVRAY_WORKSPACE_DIRdatasetsDirectory containing CSV datasets
DEVRAY_ARTIFACTS_DIRartifactsDirectory for saved model artifacts

MCP client configuration

Add the server to your MCP-compatible client. Replace C:\MLData with your own paths.

1. OpenAI Codex (CLI & VS Code)

DevRay DataLab MCP connects to the OpenAI Codex CLI and its VS Code extension through local stdio execution. Codex and its VS Code extension share configuration, so registering the server once makes it available in both.

Prerequisites

  • Install the Codex CLI (codex on PATH).
  • Install uv (provides uvx).
  • Create the dataset and artifact directories before registering:
New-Item -ItemType Directory -Force "C:\MLData\datasets","C:\MLData\artifacts"
Copy-Item "examples\sample_classification.csv" "C:\MLData\datasets\"

Register the server

codex mcp add devray-datalab `
  --env DEVRAY_WORKSPACE_DIR=C:\MLData\datasets `
  --env DEVRAY_ARTIFACTS_DIR=C:\MLData\artifacts `
  -- uvx --from devray-datalab-mcp devray-datalab-mcp
  • DEVRAY_WORKSPACE_DIR — directory containing your CSV datasets.
  • DEVRAY_ARTIFACTS_DIR — directory where trained model artifacts are saved.
  • Replace C:\MLData with your own paths.

Verify registration

codex mcp list
codex mcp get devray-datalab

Expected output shows enabled: true, transport: stdio, command: uvx, and the devray-datalab-mcp args.

Manual config.toml configuration

Codex also reads ~/.codex/config.toml (Windows: %USERPROFILE%\.codex\config.toml):

[mcp_servers.devray-datalab]
command = "uvx"
args = ["--from", "devray-datalab-mcp", "devray-datalab-mcp"]

[mcp_servers.devray-datalab.env]
DEVRAY_WORKSPACE_DIR = "C:/MLData/datasets"
DEVRAY_ARTIFACTS_DIR = "C:/MLData/artifacts"

Codex CLI and its VS Code extension share this configuration. Register once, use everywhere.

2. Cursor

Cursor uses the standard MCP client configuration format.

Add to your Cursor MCP settings (settings.json):

{
  "mcp.servers": {
    "devray-datalab": {
      "command": "devray-datalab-mcp",
      "args": [],
      "env": {
        "DEVRAY_WORKSPACE_DIR": "C:\\MLData\\datasets",
        "DEVRAY_ARTIFACTS_DIR": "C:\\MLData\\artifacts"
      }
    }
  }
}

Replace C:\MLData with your own paths. Save the file and restart Cursor for changes to take effect.

3. Claude Desktop

Add to your Claude Desktop MCP settings (claude_desktop_config.json, typically located at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS or %APPDATA%\Claude\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "devray-datalab": {
      "command": "devray-datalab-mcp",
      "args": [],
      "env": {
        "DEVRAY_WORKSPACE_DIR": "C:\\MLData\\datasets",
        "DEVRAY_ARTIFACTS_DIR": "C:\\MLData\\artifacts"
      }
    }
  }
}

Replace C:\MLData with your own paths. Save the file and restart Claude Desktop for changes to take effect.

4. Cline

Add to your Cline MCP settings (settings.json):

{
  "mcp.servers": {
    "devray-datalab": {
      "command": "devray-datalab-mcp",
      "args": [],
      "env": {
        "DEVRAY_WORKSPACE_DIR": "C:\\MLData\\datasets",
        "DEVRAY_ARTIFACTS_DIR": "C:\\MLData\\artifacts"
      }
    }
  }
}

Replace C:\MLData with your own paths. Save the file and restart Cline for changes to take effect.

5. VS Code / GitHub Copilot MCP

Add to your VS Code MCP settings (settings.json):

{
  "mcp.servers": {
    "devray-datalab": {
      "command": "devray-datalab-mcp",
      "args": [],
      "env": {
        "DEVRAY_WORKSPACE_DIR": "${workspaceFolder}/datasets",
        "DEVRAY_ARTIFACTS_DIR": "${workspaceFolder}/artifacts"
      }
    }
  }
}

Other MCP clients

DevRay DataLab MCP works with any MCP-compatible client that supports stdio transport. Use devray-datalab-mcp as the command with no arguments, and set DEVRAY_WORKSPACE_DIR and DEVRAY_ARTIFACTS_DIR as environment variables.

7. Kilo Code

{
  "mcpServers": {
    "devray-datalab": {
      "command": "uvx",
      "args": [
        "--from",
        "devray-datalab-mcp",
        "devray-datalab-mcp"
      ],
      "env": {
        "DEVRAY_WORKSPACE_DIR": "C:\\MLData\\datasets",
        "DEVRAY_ARTIFACTS_DIR": "C:\\MLData\\artifacts"
      }
    }
  }
}

Available tools

ToolWhat it doesCategory
list_datasetsList CSV datasets in the workspaceDataset
inspect_datasetLoad a CSV; report types, missing values, stats, previewDataset
analyze_correlationsPearson/Spearman correlations, strongest pairsAnalysis
analyze_data_qualityMissing values, duplicates, constants, outliers, recommendationsAnalysis
detect_potential_leakageFlag potential data-leakage issues in a datasetAnalysis
analyze_feature_driftDetect feature drift between reference and current datasetsAnalysis
prepare_datasetDefine target + task type; fit preprocessing on train data onlyPreprocessing
train_baselineTrain a classification or regression modelTraining
start_training_jobSubmit a training job (non-blocking); returns a job idTraining
get_training_statusReturn the status and result of a training jobTraining
cancel_training_jobRequest cancellation of a queued or running training jobTraining
list_training_jobsList training jobs, newest firstTraining
compare_modelsCompare multiple algorithms; select best by metric (with cross-validation)Training
automl_runRun automated ML: train candidates with CV, rank by objectiveAutoML
tune_modelTune hyperparameters of a single model with cross-validationAutoML
get_model_leaderboardReturn the ranked leaderboard of an AutoML runAutoML
export_modelSave model with pipeline, metadata, and schemaModels
predictGenerate predictions from a saved local modelModels
list_experimentsList recorded training experiments, newest firstExperiments
get_experimentGet a single recorded experiment by IDExperiments
compare_experimentsCompare multiple recorded experiments by IDExperiments
list_modelsList experiments that produced a model artifactExperiments
load_model_handleLoad a previously exported model artifact by experiment ID or nameExperiments
create_split_manifestCreate a reproducible train/validation/test split manifestData Splits
validate_split_manifestValidate a stored split manifest against a datasetData Splits
list_split_manifestsList all stored split manifests, newest firstData Splits
system_infoCPU, RAM, storage, GPU detectionSystem

Supported algorithms

Classification: Logistic Regression, Decision Tree, Random Forest, Extra Trees, HistGradientBoosting, SVM, K-Nearest Neighbors.

Regression: Linear Regression, Ridge Regression, Decision Tree, Random Forest, Extra Trees, HistGradientBoosting, SVM, K-Nearest Neighbors.

Additional models (XGBoost, LightGBM) are available via the optional devray-datalab-mcp[boosting] extra.

Optional dependencies:

  • devray-datalab-mcp[boosting] — XGBoost, LightGBM
  • devray-datalab-mcp[explain] — SHAP explanations
  • devray-datalab-mcp[parquet] — Parquet file support (pyarrow)
  • devray-datalab-mcp[excel] — Excel file support (openpyxl)

What's new in v0.3.1

  • CSV delimiter detection: Automatic detection of comma, semicolon, tab, and pipe delimiters, with explicit override support.
  • Package version consistency: __version__ now correctly reports 0.3.1.
  • CSV parsing regression tests: 7 new tests covering all delimiter types including semicolon-delimited data with 21 columns.
  • Full tool compatibility: All 27 MCP tools work correctly with the new loader.

What's new in v0.3.0

  • Non-blocking training jobs: start_training_job, get_training_status, cancel_training_job, list_training_jobs — training runs off the MCP request loop with cooperative cancellation.
  • Automated Machine Learning: automl_run, tune_model, get_model_leaderboard — train candidates with CV, rank by objective, and optionally tune with Optuna.
  • More dataset formats: Parquet (.parquet) and Excel (.xlsx/.xls) via optional [parquet]/[excel] dependencies.
  • Reproducible split manifests: create_split_manifest, validate_split_manifest, list_split_manifests — stratified/group/time-ordered splits with dataset fingerprints; test set is a protected holdout.
  • Feature diagnostics: detect_potential_leakage, analyze_feature_drift — flag potential target leakage and distribution drift.

What's new in v0.2.0

  • Real cross-validation: compare_models now performs actual cross-validation (the cv parameter was previously ignored). Preprocessing is fitted independently inside each fold to prevent data leakage. Strategies: KFold, StratifiedKFold, GroupKFold, StratifiedGroupKFold, TimeSeriesSplit.
  • Richer metrics: balanced accuracy, macro/weighted F1, PR-AUC, log loss, Brier score, confusion matrix (classification); median absolute error, MAPE (regression).
  • Calibration: probability calibration with sigmoid or isotonic methods.
  • Threshold optimization: maximize F1/precision/recall or minimize a cost matrix, with optional precision/recall constraints.
  • Feature importance: permutation importance with baseline score and per-feature mean/std.
  • Experiment tracking: SQLite-backed recording of training runs (dataset fingerprint, feature schema, algorithm, hyperparameters, library versions, metrics, duration, artifacts). Tools: list_experiments, get_experiment, compare_experiments, list_models, load_model_handle.

Quickstart workflow

Using the included sample_classification.csv (target purchased):

  1. Inspect: inspect_dataset with {"name": "sample_classification.csv"}
  2. Data quality: analyze_data_quality with {"name": "sample_classification.csv"}
  3. Correlations: analyze_correlations with {"name": "sample_classification.csv", "method": "pearson"}
  4. Train & compare: compare_models with {"name": "sample_classification.csv", "target_column": "purchased", "task_type": "classification", "metric": "f1"}
  5. Export: export_model with {"artifact_name": "churn_model", "name": "sample_classification.csv", "target_column": "purchased", "task_type": "classification", "model_name": "random_forest"}
  6. Predict: predict with {"artifact_name": "churn_model", "features": {"age": 40, "income": 80000, "credit_score": 720, "category": "electronics", "region": "east"}}

A full step-by-step tutorial with copy-ready prompts is available in docs/quickstart.md.

Links

Security and privacy

All processing is local; no data leaves your machine. File access is confined to the configured workspace/artifacts directories. Path traversal is blocked, file sizes and extensions are validated, and only artifacts created by this application are loaded for prediction. No cloud hosting, no paid AI APIs, no external downloads.

License

MIT License. See LICENSE.

Reviews

No reviews yet

Be the first to review this server!