Datadata AI Integration

⌘ K

███████╗██╗  ██╗██╗██╗     ██╗     ███████╗
██╔════╝██║ ██╔╝██║██║     ██║     ██╔════╝
███████╗█████╔╝ ██║██║     ██║     ███████╗
╚════██║██╔═██╗ ██║██║     ██║     ╚════██║
███████║██║  ██╗██║███████╗███████╗███████║
╚══════╝╚═╝  ╚═╝╚═╝╚══════╝╚══════╝╚══════╝·

Two powerful ways to connect AI agents with Datadata — MCP Server for direct protocol-level integration, and Skills for CLI-based agent tooling. Both give AI agents the ability to search, query, and manage data through natural language.

🚀 Quick Setup

Send the following message to your AI agent — it will automatically install the CLI, configure the MCP Server connection, and install all Skills in one go.

Please read https://www.datadata.com/ai-agent-setup.en-US.md and follow the steps to configure Datadata platform AI integration for me.

Your agent will guide you through OAuth authorization and skill installation. Just follow the prompts.

💡 If auto-configuration doesn't work for your agent, follow the manual steps below to integrate (MCP Server + Skills installation).

MCP Server

Datadata provides a standard MCP (Model Context Protocol) endpoint. Connect directly from Claude Desktop, VS Code, Cursor, or any MCP-compatible client. Two authentication methods are supported:

https://www.datadata.com/api/mcp/v1

OAuth 2.0

recommended

Select OAuth when configuring your MCP client — your browser will redirect to Datadata for authorization. This is the most standard and convenient method. Once authorized, the agent can call all MCP tools without manually managing API Keys.

API Key

You can also authenticate via API Key — see the Authentication section below for permission configuration.

Skills

datadata-manual

skill

Interactive operations guide for the Datadata platform — search datasources, explore table structures, execute queries, manage Data Spaces, and schedule queries. Covers all day-to-day Datadata platform capabilities via MCP Server. For generating standalone scripts (crawler/ETL/batch), use datadata-rest-api skill instead.

Core Capabilities

Search datasources — public & private search by username/keyword
Metadata queries — inspect datasource info, list tables, describe column schemas
Metadata enhancement — set table & column comments for better data understanding
Execute SQL/DQL queries — DuckDB & ClickHouse engines
Data Spaces management — create tables, batch insert rows, drop tables
Schema scanning — trigger async scans to refresh table metadata
Scheduled queries — periodic query execution & result delivery

Triggers:

Datadatadata explorationSQL querytable inspectionexecution resultsdata analysiscreate tablewrite datascheduled tasks

datadata-dql

skill

Write DQL (Datadata Query Language) scripts — a Starlark-based scripting language for data transformation, cleaning, generation, and custom processing logic.

Core Capabilities

DataFrame/Series operations — transformation, cleaning, aggregation, merging, anomaly detection
SQL queries — execute SQL via query() and return DataFrame
HTTP requests — call external APIs via fetch()
2D drawing — Canvas basic graphics
Built-in modules — JSON, Time, Math, and more

Triggers:

Data transformationdata cleaningdata generationcustom data processing logic

datadata-memory

skill

Manage AI persistent memory via Datadata MCP Server — add, search, update, and delete memories with semantic search and multi-dimensional filtering. Store user preferences, project conventions, and domain knowledge across sessions.

Memory Management

Add memories — store atomic facts with tags, categories, and metadata
Semantic search — keyword or vector-based search with multi-dimensional filters
Update & delete — modify or remove existing memories
Deduplication — auto-merge similar memories, resolve conflicts
User & agent scoping — global user memory or agent-specific context

Triggers:

remembermemoryrecalldon't forgetnote this downpreviously mentionedsearch memorydelete memory

datadata-rest-api

skill

REST API reference for Datadata — complete endpoint documentation with urllib.request examples (zero extra dependencies). Primary use case: generating standalone Python scripts for crawling, ETL, batch processing, and any scenario requiring direct Datadata API calls. For interactive operations, use datadata-manual skill instead.

REST API Reference

Complete endpoint documentation — all endpoints with request/response examples
Ready-to-use urllib.request code snippets — zero extra dependencies
Datasource operations — search, metadata, table structure, schema scanning
SQL queries — execute-adhoc (SELECT only), result download (NDJSON/CSV)
Data Spaces — create tables, batch insert rows, drop tables
Device authorization — automatic API Key acquisition/refresh in scripts

Triggers:

Datadata APIcrawler scriptETLbatch processingdata pipelineAPI integration

Installation

Install all Datadata skills with a single command:

npx skills add datadata-team/datadata-skills

View package on skills.sh→

Authentication

The skill supports two ways to authenticate:

Automatic

recommended

If no API Key is configured, the datadata-rest-api skill will automatically guide you through device authorization when you first run a command — just follow the link displayed in the terminal to complete sign-in. The key is then cached locally for 90 days.

Manual

Create an API Key in the Datadata platform with the required permissions:

Permission	Purpose
queries:execute-adhoc	Run SQL queries
executions:get	Retrieve query results
datasources:read	Read datasource metadata
datasources:scan	Trigger async schema scans
data-spaces:write	Create tables and insert data

Avatar → Settings → "API Keys" (left sidebar) → Create a new key

Select the required permissions

Set it as an environment variable:

export DATADATA_API_KEY="ak_..."