Looker October 2026 Updates: New Explore, Gemini-Powered Analytics, and MCP Server for AI Agents
In October 2026, Google introduced Looker 26.20, an update featuring a redesigned Explore interface, Gemini-powered visualization tools, and additional data analysis capabilities. Changes included a new Merge Queries workflow, AI-generated explanations of data, new chart types, and table row grouping.
Separately, on October 26, Google announced that the Looker-managed MCP Server had reached General Availability. This allows external AI agents to connect to enterprise data and LookML semantic models without deploying a separate intermediary MCP server.
The updates are available in Looker (original) and Looker (Google Cloud core), although the timing of individual features depends on each instance's version and update channel.
The New Explore Interface Becomes the Default
On October 29, Google announced that the redesigned Explore interface and the new Merge Queries workflow were generally available and enabled by default in Looker 26.20.
Explore supports self-service data analysis based on LookML models. Users select measures, dimensions, filters, and visualizations, and Looker then generates a query against the connected data source.
The new interface centers on a Configure panel with three sections: All Fields, Fields in Use, and Style.
All Fields contains the available measures and dimensions. Fields in Use displays the elements of the current query and the selected visualization. Style contains formatting settings.
Field search has also been redesigned. Users can find fields by name, description, and related groups, making large data models easier to work with.
The interface also provides an Explore overview, including the number of available fields and a link to the associated LookML model for users with the required permissions.
These changes are primarily intended for business users who need to explore data independently without constantly relying on reporting developers.
Updated Merge Queries for Combining Analytics Data
A new Merge Queries interface also became generally available in Looker 26.20.
The tool combines the results of multiple Explore queries using shared dimensions.
For example, a marketing team can compare advertising spend with sales metrics when those datasets reside in different Looker models and have appropriate common fields.
The redesigned workflow has three panels: a list of queries being merged, field selection, and a preview of the result.
Users can edit source queries without leaving the merging workflow.
The system also suggests suitable fields for joining data, such as date, region, or other shared dimensions.
To reuse queries, users can add an existing query by its URL.
However, Merge Queries does not replace proper data modeling. If metrics have different levels of granularity, such as campaign-level advertising spend and order-level sales, an incorrect join can duplicate rows and distort aggregated values.
For ongoing enterprise reporting, complex joins are better implemented in the data warehouse or LookML, where keys and aggregation rules can be managed centrally.
Gemini Explore Summaries: Written Explanations of Data
On October 26, Google made Gemini Explore Summaries generally available for Looker 26.20 instances.
The feature uses Gemini to generate written descriptions of Explore results.
Instead of interpreting every chart independently, users receive a summary that helps explain the main characteristics of the selected data.
For example, when analyzing regional sales trends, Gemini can describe differences between regions and the most notable changes.
This is useful for people who work with reports regularly but do not specialize in data analysis.
Gemini does not, however, receive unrestricted access to all enterprise information. Its work is limited by the available Explore context and the configuration of the relevant Looker instance.
Generated explanations must be reviewed: an AI model may accurately describe a metric's movement while incorrectly interpreting what caused it.
For example, an increase in advertising spend occurring alongside higher sales does not, by itself, establish a causal impact from advertising.
Visualization Assistant: Configuring Charts with Natural Language
On October 29, Google announced the General Availability of Visualization Assistant.
The tool uses Gemini to adjust visualizations based on natural-language prompts.
Users can describe how they want data displayed instead of manually locating the right options in the chart editor.
For example, an analyst can request a different number format, adjustments to axis styling, or changes to other supported visualization settings.
Visualization Assistant translates the user's instructions into chart configuration settings.
To use it, an administrator must enable Gemini in Looker and Looker Assistants in the relevant platform administration settings.
Visualization Assistant should not be confused with generating new data. It is designed to configure the visual presentation of existing results, not change metric values or automatically fix errors in the underlying model.
AI-assisted chart configuration therefore does not eliminate the need to validate dimensions, aggregations, and filters.
KPI Charts and Bullet Charts for Tracking Targets
New performance metric visualization tools became generally available in Looker 26.20.
KPI (Single Value) replaces the previous Single Value visualization and offers more formatting options for key metrics.
The format is designed to display individual metrics such as revenue, order volume, lead count, or customer acquisition cost.
For marketing dashboards, it is especially useful for creating an executive-level overview with several core KPIs.
Bullet Chart compares an actual value against a defined target.
The chart displays the main metric, a target line, and background ranges representing different performance levels.
For example, a company can display actual ROAS against its target and define additional ranges for below-target, on-target, and above-target results.
Unlike a standard numeric indicator, a Bullet Chart immediately shows where a metric stands relative to the established thresholds.
Both visualization types became generally available in the October update.
Table Row Grouping and New Histograms
On October 29, Google also made Table Row Grouping generally available.
The feature groups table rows into a hierarchy and lets users control how individual groups are displayed.
For example, a marketing team can organize results first by country, then by advertising platform, and then by campaign.
This makes large tables with repeated dimension values easier to read.
Row grouping changes how results are displayed, allowing reports to use a clearer structure.
Looker also introduced a new Histogram visualization type.
A histogram shows the distribution of numeric values across intervals.
For example, it can be used to examine the distribution of average order values, user session lengths, or customer acquisition costs.
Unlike a standard bar chart organized by categories, a histogram groups continuous numeric values into intervals and shows the frequency of observations.
For marketing analytics, this chart can be more informative than an average when the goal is to understand variability or identify outliers.
Dashboard Tile Groups: Organizing Dashboard Elements
The October release introduced Dashboard Tile Groups in Preview.
The feature allows multiple dashboard elements to be combined into collapsible groups.
For example, a single screen can contain separate sections for Paid Search, Paid Social, Direct Marketing, and Revenue Analytics.
Each section can contain relevant tables, charts, and KPIs.
A dashboard editor can move, duplicate, and delete multiple elements as a group.
The feature also supports defining groups in LookML.
For complex enterprise dashboards, this can reduce visual clutter and organize metrics by business function.
Dashboard Tile Groups remains in Preview, however, unlike the new Explore, KPI Chart, and Table Row Grouping, which have reached GA.
Looker-Managed MCP Server Reaches General Availability
On October 26, Google announced the General Availability of Looker's built-in MCP server.
Model Context Protocol allows external AI agents to interact with data and tools in connected systems.
The Looker-managed MCP Server runs within Looker's own infrastructure, so companies do not have to deploy and maintain a separate intermediary service.
Compatible AI clients can connect to the MCP server, including Gemini CLI, Claude Desktop, Cursor, and Copilot.
Through the available tools, an agent can interact with Looker data and LookML semantic models.
For example, an analyst can use a connected AI assistant to retrieve specific sales metrics or explore data available through the company's Looker model.
A key distinction from direct AI access to a database is the use of the existing LookML semantic layer.
This preserves centrally defined metrics, dimensions, and business logic.
For example, if LookML specifies which orders count toward Net Revenue, the AI tool can use that defined metric rather than inventing a formula based on table names.
The accuracy of the result still depends on the quality of the underlying LookML model.
Managing AI Agent Access with OAuth
Alongside the MCP server, Google introduced an administrative Manage OAuth Clients interface.
It is used to register external AI applications and manage their access to Looker.
An administrator can select a preconfigured integration from the Application Library or register a custom OAuth client.
Custom integrations support Redirect URI settings and group-based access restrictions.
An AI agent runs queries within the permissions of the authenticated user.
If an employee cannot access a particular model, Explore, or dataset, the connected agent should not be able to access it either.
MCP server tools are disabled by default. An administrator must explicitly allow the required tools.
The built-in server requires a Looker-hosted instance of Looker (original) or Looker (Google Cloud core).
For self-hosted instances, Google offers an alternative approach using MCP Toolbox for Databases.
Therefore, the GA release of MCP does not automatically make all enterprise data available to every AI application. Access depends on server settings, OAuth client configuration, and Looker's existing permissions.
Data Lineage Reaches GA in Looker Google Cloud Core
On October 26, Google also announced the General Availability of Data Lineage for Looker (Google Cloud core).
Data Lineage helps users examine where data originates and how source systems, models, and analytics assets are connected.
The update supports lineage tracking for LookML Dashboards and Conversational Analytics Agents.
For example, when investigating a business metric, an analyst can identify the data sources and models used to build the corresponding report.
This is useful when changing the data warehouse structure or evaluating the effects of table updates.
If a developer plans to change a field used by multiple enterprise dashboards, dependency information helps assess the potential impact.
Google also separated Data Lineage administration from internal Knowledge Catalog settings: it is enabled through the Data Lineage API and associated metadata collection controls.
Unlike the MCP Server, which is available in both supported cloud versions of Looker, this Data Lineage update applies specifically to Looker (Google Cloud core).
Database Connection Diagnostics
Looker 26.20 added tools for diagnosing connections to data sources.
When a connection test fails, Looker automatically checks DNS resolution and TCP connectivity.
The results of those checks are added to the error message.
Administrators can also use Test Reachability on the Connections page to check whether a specified address and port are reachable.
For example, if Looker cannot connect to a corporate database, the new diagnostics help distinguish a network connectivity problem from a connection configuration error.
This reduces the amount of manual troubleshooting required during technical incidents.
A successful network check, however, does not establish that the credentials are correct or that the user has permission to run SQL queries.
The tool primarily diagnoses network connectivity rather than every possible cause of database errors.
Looker 26.20 Availability and Release Channel Differences
The availability of new capabilities depends on the installed Looker version and its update channel.
Google Cloud offers several release channels, including Regular, Rapid, and No Channel.
As of October 29, 2026, Looker (Google Cloud core) version 26.20 was available through Rapid and No Channel, while Regular remained on version 26.18.
This means that even after a feature is officially declared generally available, it may still be missing from a Looker instance that has not received the relevant update.
Administrators should verify the installed version and the settings for individual AI features.
MCP also requires enabling the relevant tools and configuring OAuth clients.
Visualization Assistant requires Gemini in Looker and Looker Assistants to be enabled.
Feature availability is therefore determined not only by licensing configuration but also by platform version and administrative settings.
Looker: Analytics Through a Visual Interface and AI Agents
Looker is Google Cloud's enterprise business intelligence platform. It uses LookML to centrally define business metrics and provide governed access to data.
The October updates advance two major areas of the product.
The first is self-service analytics within Looker. The new Explore, Merge Queries, Gemini Summaries, and Visualization Assistant simplify querying, interpreting results, and preparing visualizations.
The second is access to Looker's semantic layer from external AI applications through the MCP Server.
For ecommerce businesses, these tools can support analysis of orders, categories, customers, and marketing performance. For service businesses, they can support the study of leads, conversions, and operational metrics. For B2B companies, they can support work with pipeline, CRM, and revenue reporting.
The MCP Server reaching GA is particularly important because it enables compatible AI assistants to use corporate metric definitions from LookML while retaining existing access restrictions.
AI features still do not replace a reliable data model. The accuracy of answers, charts, and analytical conclusions depends on source data quality, correct LookML definitions, and selected query parameters.
Sources
- https://docs.cloud.google.com/looker/docs/release-notes
- https://cloud.google.com/blog/products/business-intelligence/looker-explore-ai-and-interface-updates/
- https://docs.cloud.google.com/looker/docs/2620/mcp
- https://docs.cloud.google.com/looker/docs/2620/release-changelog
- https://www.datastudios.org/post/looker-26-20-new-explore-experience-default-gemini-features-ga



