Dataiku Cobuild
AI agent · Data & analytics agent · Data & analytics by Dataiku
The natural-language way to build and oversee governed enterprise AI.
Dataiku Cobuild is an agentic interface that enables users to build end-to-end AI projects, from data pipelines to ML models, using natural language. It empowers both business experts and developers to accelerate AI delivery while maintaining full human oversight through visual, editable workflows.
Agent specs
The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.
- Languages
- English, German, Japanese, French, Spanish, Italian
Category: AI Agents
- Channels
- In-app (Dataiku platform), External IDEs
- Interface Type
- Natural Language / Chat-based
- Output Formats
- Visual Flows, ML Models, Web Apps, DOCX, PPTX, PDF
- Developer Tool Integrations
- Claude Code, Codex, Cursor, Snowflake CoCo
- Platform Integration
- Dataiku Catalog, Agent Hub, LLM Mesh
- Governance Model
- Human-in-the-loop with visual approval workflows
- Natural language interface for building governed AI projects
- Generates end-to-end data pipelines, ML models, and web apps
- Delivers results as visual, editable Flows for human oversight
- Integrates with Claude Code, Cursor, and Codex via Dataiku Headless
- Built-in integration with the Dataiku Catalog for trusted assets
Overview
Dataiku Cobuild is an agentic interface designed for both technical and non-technical business users to build and manage AI projects using natural language. It solves the bottleneck of AI creation by allowing users to describe business objectives in plain language, which the agent then transforms into complete, governed AI projects including data pipelines and machine learning models. The agent operates across the entire Dataiku platform, providing a unified way to explore data, generate automation agents, and oversee existing agent populations. What sets it apart is its delivery of results as visual, editable Flows, ensuring that speed is balanced with human oversight and centralized governance. Additionally, through Dataiku Headless, it extends these capabilities into popular developer tools like Cursor and Claude Code, allowing for seamless migration of work and administrative automation.
Core Agentic Capabilities
- Natural Language Project Creation: The agent transforms business objectives described in plain language into complete AI projects within the Dataiku platform.
- End-to-End Generation: It generates end-to-end projects including data pipelines, machine learning models, agents, and web applications.
- Data Exploration: Users can explore data and get trusted answers using natural language across the whole platform.
- Agent Oversight: The agent provides an interface to oversee and manage the population of agents running across the business.
Governance and Control
- Visual Flow Output: All generated work is delivered as a visual Flow that teams can inspect, edit, and approve.
- Catalog Integration Building: The agent is integrated with the Dataiku Catalog to find and build upon trusted assets like tools, models, and data.
- Governed AI Construction: It allows for the creation of governed AI, ensuring speed does not come at the expense of control.
- Interactive Refinement: Users can refine documents and artifacts generated by agents simply by using chat.
Developer Experience
- Headless Coding Support: Dataiku Headless brings Cobuild capabilities into external coding tools like Claude Code, Codex, Cursor, and Snowflake CoCo.
- Context-Free Migration: Developers can create Dataiku projects and migrate work developed in other environments without switching context.
- Administrative Automation: The agent can perform administrative tasks within the Dataiku environment through external developer tools.
- Developer Governance Extension: It supports the creation of trusted AI within the developer's preferred coding environment while remaining under Dataiku governance.
What it does
- Transform plain-language objectives into AI projects
- Generate data pipelines
- Build machine learning models
- Create automation agents and web applications
- Explore data and answer questions
- Oversee running agent populations
- Migrate work from external coding tools
- Perform administrative tasks
Use cases
- Business analysts use natural language to generate data preparation pipelines and SQL queries without writing code.
- Data scientists use Cobuild to rapidly prototype end-to-end machine learning projects from a simple goal description.
- Developers use Dataiku Headless to create and manage Dataiku projects directly from within coding tools like Cursor or Claude Code.
- Governance teams oversee and manage the lifecycle of various AI agents running across the business from a single interface.
- Users refine agent-generated documents and artifacts through interactive chat before exporting them as professional reports.
Specifications
- Channels
- In-app (Dataiku platform), External IDEs
- Integrations
- Claude Code, Codex, Cursor, Snowflake CoCo, Dataiku Catalog, Agent Hub
- Human control
- Generated projects are delivered as visual Flows for human inspection, editing, and approval before deployment.
- Deployment
- Cloud, On-premises, Air-gapped (as part of the Dataiku platform)
- Languages
- English, German, Japanese (based on available platform pages)
- Interface Type
- Natural Language / Chat-based
- Output Formats
- Visual Flows, ML Models, Web Apps, DOCX, PPTX, PDF
- Developer Tool Integrations
- Claude Code, Codex, Cursor, Snowflake CoCo
- Platform Integration
- Dataiku Catalog, Agent Hub, LLM Mesh
- Governance Model
- Human-in-the-loop with visual approval workflows
How it works
- Describe Objective: The user describes a business goal or data transformation requirement in plain, natural language.
- Generate Project Components: Cobuild generates the necessary data pipelines, machine learning models, and automation steps to achieve the goal.
- Review and Refine: The generated work is presented as a visual Flow within Dataiku, where teams can inspect every step and make manual edits.
- Approve and Deploy: Once validated, the project is approved and deployed within the governed Dataiku environment.
FAQ
What is Dataiku Cobuild?
Dataiku Cobuild is an agentic interface that allows users to explore data, build governed AI, and oversee agents using natural language across the entire platform.
What kind of projects can Cobuild create?
It generates complete end-to-end projects including data pipelines, machine learning models, agents, and web applications, all delivered as editable visual Flows.
How does Cobuild work with developer tools?
Dataiku Headless extends Cobuild to external coding tools like Claude Code, Cursor, and Codex, allowing developers to build and manage Dataiku projects without leaving their IDE.
Can Cobuild use existing company data and models?
Yes, Cobuild is integrated with the Dataiku Catalog, allowing it to find and build upon trusted assets like existing tools, models, and data.
Added by AntSpace from public sources (dataiku.com).