convez.ai
MCP Server
datarobot-community

af-component-datarobot-mcp

datarobot-community/af-component-datarobot-mcp

Ver no GitHub
BomPluginScore · de 0 a 100

DataRobot MCP AF Component for building Model Context Protocol servers with deep DataRobot integration.

Instalação

$npx convez-ai datarobot-community/af-component-datarobot-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
3
Acompanhando
2
Saúde
87%
Último commit
há 2 dias
Versão
0.0.88
Issues abertas
1
Contribuidores
12

README

A tradução para o português deste README ainda está na fila.

This component is part of the DataRobot App Framework and deploys a DataRobot MCP (Model Context Protocol) server as a DataRobot custom model application. It is designed for app developers and platform engineers who need to expose DataRobot ML/AI capabilities to LLM-based agents and tools.

The component ships a ready-to-deploy MCP server that includes a comprehensive set of DataRobot predictive tools (project management, model training, deployments, and batch/real-time predictions) as well as integrations for popular collaboration platforms (Google Drive, Jira, Confluence, and Microsoft Graph). It runs as a FastMCP server and can be applied multiple times in the same project under different names to support one-to-many MCP backends in a single template.

Table of contents

Prerequisites

The following tools must be installed before using this component.

The base component must be applied to the project before this component. See Component dependencies for details.

Quick start

Run the following command in your project directory:

dr component add https://github.com/datarobot-community/af-component-datarobot-mcp .

If you need additional control, you can run this to use copier directly:

uvx copier copy datarobot-community/af-component-datarobot-mcp .

During setup you will be prompted for an mcp_app_name. This name scopes all generated files and answers for this MCP backend, which lets you apply the component multiple times in the same project for different server instances.

Component dependencies

Required

The following components must be applied to the project before this component:

NameRepositoryRepeatable
basehttps://github.com/datarobot-community/af-component-baseNo

Local development

After applying the component, the following paths contain the key files for development.

PathPurpose
template/{{mcp_app_name_file}}/Generated MCP server source (tools, server entrypoint).
template/docs/datarobot-mcp/MCP server documentation (rendered to docs/datarobot-mcp/ in the target project).
.datarobot/answers/drmcp-{{ mcp_app_name }}.ymlCopier answers file for this instance.

On dr component add and dr component update, copier runs uv lock in {mcp_app_name}/ so uv.lock matches your chosen app name and current dependencies.

To run the MCP server locally:

uv run python -m MCP_APP_NAME

Refer to the MCP server documentation for the full developer guide, including OAuth provider configuration for integration tools.

Documentation

After applying this component, MCP guides are available under docs/datarobot-mcp/ in your project. In this repository, the template sources live under template/docs/datarobot-mcp/:

DocumentDescription
OverviewGetting started, local dev, and deployment overview
MCP client setupConfigure Cursor, VS Code, and Claude Desktop
Server architectureProject structure and configuration reference
Dynamic tool registrationTurn DataRobot deployments into tools automatically
Custom toolsAuthor domain-specific tools
Deployment info toolsQuery deployment features and build prediction datasets

Updating

All components should be regularly updated to pick up bug fixes, new features, and compatibility with the latest DataRobot App Framework.

For automatic updates to the latest version, run the following command in your project directory:

dr component update .datarobot/answers/drmcp-<mcp_app_name>.yml

If you need more fine grained control and prefer using copier directly, you can run this to have more control over the process:

uvx copier update -a .datarobot/answers/drmcp-<mcp_app_name>.yml -A

Available tools

The MCP server provides tools organized into DataRobot platform tools and third-party integration tools (data connectors and web search).

Data Connectors — Confluence

  • confluence_get_page—Retrieve Confluence page contents by ID or by exact title within a space.
  • confluence_create_page—Create a new Confluence page in a space, optionally under a parent page.
  • confluence_add_comment—Add a comment to an existing Confluence page.
  • confluence_search_space—Search Confluence content using Confluence Query Language (CQL).
  • confluence_update_page—Update the body of an existing Confluence page (requires current page version number).

Data Connectors — Jira

  • jira_search_issues—Find Jira issues matching a JQL query.
  • jira_get_issue—Get full details for a single Jira issue (requires issue key, e.g., PROJ-123).
  • jira_create_issue—Create a new Jira issue in a project.
  • jira_update_issue—Update fields on an existing Jira issue.
  • jira_transition_issue—Move a Jira issue to a new workflow status.

Data Connectors — Google Drive

  • gdrive_find_contents—Search Google Drive for files and folders, using optional query and pagination parameters.
  • gdrive_read_and_export_content—Read and export the text content of a Google Drive file.
  • gdrive_create_file—Create a new file or folder in Google Drive.
  • gdrive_update_metadata—Rename, star, or delete a Google Drive file.
  • gdrive_manage_access—Add, update, or remove sharing permissions on a Google Drive file.

Data Connectors — Microsoft 365

  • microsoft_graph_search_content—Search SharePoint and OneDrive for files and list items.
  • microsoft_graph_share_item—Share a SharePoint or OneDrive file or folder.
  • microsoft_graph_create_file—Create a new plain-text file in OneDrive or SharePoint.
  • microsoft_graph_update_metadata—Update name or metadata on a SharePoint list item or drive item.

Web Search — Perplexity

  • perplexity_search—Search the public web and return ranked sources and snippets for a question.
  • perplexity_sonar—Get a cited AI-generated answer or research report from Perplexity; available model options are sonar, sonar-reasoning-pro, and sonar-deep-research.

Web Search — Tavily

  • tavily_search_web—Search the public web by keyword and return ranked results.
  • tavily_extract_text—Extract clean text from one or more web page URLs.
  • tavily_list_links—List links discovered under a website URL.
  • tavily_crawl_site—Crawl a website to collect multiple related pages.

DataRobot Documentation

  • datarobot_docs_fetch_page—Retrieve the full text of a DataRobot documentation page by URL.

DataRobot — Use Cases

  • datarobot_usecases_list—List DataRobot Use Cases with optional name filter.
  • usecases_list_assets—List datasets, deployments, and experiments linked to use cases.

DataRobot — Catalog (Datasets & Datastores)

  • catalog_upload_dataset—Upload or register a dataset from a local file or URL.
  • catalog_list_datasets—List datasets registered in DataRobot.
  • catalog_get_preview—Get metadata and optional sample rows for a dataset.
  • catalog_list_datastores—List external database and storage connections configured in DataRobot.
  • catalog_browse_datastore—Browse schemas, tables, or folders inside an external datastore.
  • catalog_query_datastore—Run SQL against an external datastore.
  • catalog_check_timeseries_eligibility—Check whether a dataset is ready for time series modeling.
  • catalog_analyze_dataset—Return profile of an AI Catalog dataset (columns, types, and missing values).
  • catalog_suggest_ml_problems—Recommend a target, which defines a problem type, from a dataset.
  • catalog_get_eda_insights—Run exploratory data analysis (EDA) on a dataset and return summary statistics.

DataRobot — Modeling (Projects & Models)

  • models_get_bestmodel—Get the top-performing model on a model leaderboard.
  • modeling_score_dataset—Score a catalog dataset with a trained model from a project leaderboard (async job).
  • modeling_list_models—List all leaderboard models.
  • modeling_get_modeldetails—Get training metrics and diagnostics for a leaderboard model.
  • modeling_list_projects—List all modeling projects associated with the user's organization.
  • modeling_get_project_dataset—Find a project dataset (requires project name).
  • modeling_start_autopilot—Start or resume Autopilot training for a modeling project.
  • modeling_get_model_roc—Get ROC curve data—classification, performance, and statistics—for a binary classification model.
  • modeling_get_model_feature_impact—Get model feature impact, which identifies the features most strongly driving model decisions.
  • modeling_get_model_lift_chart—Get lift chart data for a binary or multiclass classification model.

DataRobot — Deployments

  • deployment_get_list—List MLOps deployments in DataRobot.
  • deployment_get_model_info—Get the model record linked to a deployment (requires deployment ID).
  • deployment_create_deployment—Create a new deployment from a trained leaderboard model.
  • deployment_get_prediction_history—Retrieve logged prediction history for a deployment.
  • deployment_get_info—Get deployment scoring information: feature details, model settings, and deployment ID.
  • deployment_generate_prediction_sample—Generate sample prediction rows with required columns for valid prediction output.
  • deployment_validate_prediction_data—Validate inline CSV prediction data against a deployment schema.
  • deployment_get_features—Get input features, including target summary, for a deployment.

DataRobot — Predictions

  • predict_batch_predictions_from_dataset—Generate batch predictions from a deployment using a supplied dataset.
  • predict_batch_predictions_from_holdout_data—Generate batch predictions from a deployment using the holdout partition data.
  • predict_get_batch_job_status—Check status of a batch prediction job.
  • predict_get_batch_results—Download scored results from a completed batch prediction job.
  • predict_score_catalog_realtime—Score a catalog dataset through a deployment and return rows immediately.
  • predict_score_inline_realtime—Score inline CSV or JSON rows through a deployment immediately.

DataRobot — Vector Databases (VDB)

  • vdb_list—List deployed vector database (VDB) deployments.
  • vdb_query—Run semantic search against a vector database deployment.

Integration tools (Google Drive, Jira, Confluence, Microsoft 365) require OAuth authentication configured via DataRobot OAuth providers. Perplexity and Tavily tools require API keys. See MCP client setup and server architecture for configuration details.

Troubleshooting

Common issues and their solutions are listed below.

dr command not found — Ensure the DataRobot CLI is installed and on your PATH. See the CLI docs for install instructions.

Copier prompts fail or produce unexpected output — Confirm you are running uv 0.4+ and that copier resolves via uvx. Run uvx copier --version to verify.

Integration tools return auth errors — OAuth providers must be configured in DataRobot before integration tools (Google Drive, Jira, Confluence, Microsoft Graph) will work. See MCP client setup.

Multiple instances conflict — Each instance must use a unique mcp_app_name. If two instances share a name, their answers files and generated directories will collide.

For additional help, open an issue on the GitHub repository or contact DataRobot support.

Next steps and cross-links

Explore the following resources to learn more or extend this component.

Contributing, changelog, support, and legal

Contributing — Fork the repository, make changes on a feature branch, ensure task lint passes, and open a pull request. See CONTRIBUTING.md if present.

Changelog — See CHANGELOG.md for version history. This component follows semantic versioning; the current version badge at the top of this README reflects the latest release tag.

Getting help — Open an issue on the GitHub repository or reach out via DataRobot support.

License — This project is licensed under the terms shown in the LICENSE file.

Ver o original no GitHub

Bom para

Relacionados