Next upAI x Bio Pitch Contest
News

Dataiku announces cross-platform Agent Management and expands Cobuild

Dataiku announced a standalone oversight product spanning enterprise AI agents on multiple vendor platforms, while extending Cobuild into governed analysis, coding workflows, and agent-portfolio management.

D
Sep 27, 2026 · 2 min read

Dataiku announced Agent Management, a standalone product designed to inventory, monitor, and govern enterprise AI agents across multiple vendors’ platforms. The product is due to become generally available in October 2026. Dataiku says access will be priced annually per instance, with monitoring charged per agent, but it did not disclose an exact release date or prices.

Agent Management is designed to give organizations a consolidated record of agents running in Dataiku and external systems without requiring those agents to move onto one execution platform. Dataiku describes it as an oversight layer rather than an AI gateway: the product observes and records agents but does not operate them or intervene in their decisions.

According to Dataiku, Agent Management connects to AWS Bedrock, Databricks Agents, Google Vertex, Microsoft Copilot Studio and Azure Foundry, Salesforce Agentforce, Snowflake Cortex, and Dataiku. It also supports OpenTelemetry for custom environments. The depth of each integration depends on the data the connected platform exposes.

The product scans connected platforms into a central inventory that can include each agent’s owner, purpose, dependencies, and recent activity. Dataiku says it can track usage, cost, quality, health, behavior, and business value over time, then flag drift or unusual patterns for human review. The available evidence does not independently test those monitoring and detection claims.

For higher-risk agents, Dataiku says teams can document checks, record named risks and mitigations, certify an agent, and rerun tests on a schedule to maintain an audit trail. The company has not specified which of those functions will be available for every supported third-party platform.

Dataiku also expanded Cobuild, its natural-language interface, beyond no-code project generation. The additions cover governed data analysis, developer workflows, agent-portfolio oversight, and a broader catalog of approved AI assets. Dataiku said they will roll out through fall 2026, without providing feature-by-feature dates.

Cobuild Insights is designed for technical and nontechnical users to ask questions in plain language and receive answers grounded in governed data. Dataiku says queries respect existing access rights and are logged for audit. Users can then continue the same thread in deeper Dataiku workflows.

A separate capability, Dataiku Headless, exposes Cobuild through coding tools including Claude Code, OpenAI Codex, and Cursor. Dataiku says developers can use it to build or modify pipelines, models, dashboards, and agents, as well as migrate legacy analytics into inspectable Dataiku projects.

Inside Agent Management, Dataiku says Cobuild can recommend business-value metrics and help diagnose and remediate risk alerts across an agent portfolio. Cobuild’s expanded AI Catalog covers vetted datasets, models, agents, and semantic models, and can assemble selected assets into an answer or a visual pipeline with lineage. Cobuild itself has been generally available since June 18, 2026.

More news