Dataiku has launched Dataiku Headless, a new capability that lets developers build, migrate and manage governed AI and data projects directly from coding tools such as Claude Code, Codex and Cursor. The company unveiled it at Succeed, its annual flagship conference in New York, with a Singapore dateline marking the regional announcement.
Key Takeaways
- Dataiku Headless brings the company’s natural-language agentic interface, Cobuild, into coding tools developers already use, including Claude Code, Codex and Cursor.
- It lets developers build governed Dataiku projects from a coding agent, migrate legacy workflows such as Excel or Alteryx files, and diagnose failing jobs across environments.
- Dataiku Headless is available now as a plugin on the Claude Code and Codex marketplaces, and is compatible with Snowflake CoCo, GitHub Copilot and AWS Kiro.
Why Dataiku built it
Coding agents like Claude Code and Codex have sped up how technical teams build software, but Dataiku argues that speed has created a new bottleneck: once something is built, few people beyond the developer who built it can actually see it, review it or run it safely in production. Headless is designed to close that gap without pulling developers out of the coding agent they already use day to day.
Under the hood, Headless acts as a control surface: a coding agent uses it to inspect and operate Dataiku, calling on Cobuild, Dataiku’s existing natural-language agentic interface, whenever a task requires creating or changing a project.
What developers can do with it
Dataiku lists three main uses for Headless. Developers can build directly from a coding agent, which delegates the work to Cobuild to create a visual flow, datasets, transformations, models or custom agents that the wider team can inspect. They can migrate legacy workflows, pointing a coding agent at something built outside Dataiku, such as an Excel workbook or Alteryx pipeline, and having it rebuilt as a governed, visual Dataiku project, a process Dataiku says can run hundreds or thousands of times for large modernisation projects. And administrators can use it to investigate and support Dataiku environments, asking a coding agent to find jobs that have failed across multiple projects and remediate them by changing project or admin settings.
“Developers aren’t going to stop working in Claude Code or Codex, and we don’t think they should have to. What they need is a way to turn that speed into something the rest of the business can actually see, review, and run in production. That’s what Dataiku Headless does — it meets technical teams in the tools they already trust, and hands back a Dataiku project instead of a script only they can read.”
Jed Dougherty, SVP of AI & Platform, Dataiku
Availability
Dataiku Headless is available now as a plugin on the Claude Code and Codex marketplaces, and is compatible with Cursor, Snowflake CoCo, GitHub Copilot, AWS Kiro and other agentic coding tools.
