On the occasion of its event which opens this June 1 in San Francisco, the big data specialist lifts the veil on a series of new products. Objective: facilitate mass data management.
Snowflake is organizing its 2026 summit from June 1 to 4 at the Moscone Center in San Francisco. During the opening keynote of the event, the big data expert lifted the veil on a series of announcements. The first targets its platform targeting agent management and called Snowflake Cowork (formerly Snowflake Intelligence). “Now, Snowflake Cowork has connectors to interface with your application ecosystem. These connectors support, for example, Slack, Google Drive, Salesforce or Workday,” explains Baris Gueltekin, head of AI at Snowflake. “Our philosophy is to bring AI and data together, not the other way around. Our platform taps into data where it is (editor’s note without making a copy). This allows us to generate the layers of context essential to agents in an efficient manner.”
Under the hood, Snowflake aims to be model agnostic. The actor has partnerships with Anthropic, Google and OpenAI (90% of the market). Agreements which allow it to include new versions of LLMs from these players as soon as they are launched. “When you think about how quickly the industry changes, being able to adapt quickly and use the best model as soon as it becomes available becomes a key value proposition,” insists Baris Gueltekin. “When you’re deploying AI across an organization, you need to make sure you’ve thought about your governance from the start. By being close to the data, we deliver that in a very powerful way by making sure only the person who has access to the data can see it.”
Cortex Sense to grasp context
Snowflake Cowork now supports skills securely. “This means that they run on the server. With this in mind, we have set up a sandbox in which all the code can run in isolation,” announces Baris Gueltekin. To best understand the context in which agents operate, Snowflake presented another new feature. Called Cortex Sense, the application in question makes it possible to automate the links between data and the business activity of user companies.
“Cortex Training is designed to support companies that build, refine and optimize models in a personalized way”
“We are also launching a product called Cortex Training. It is designed to support companies that build, refine and optimize models in a personalized way,” adds Baris Gueltekin. “Concretely, it is a managed training cloud space which provides access to GPUs but also to services allowing the management and execution of the training phases of LLMs on Snowflake.” To secure the execution of agents, an identity management layer is made available. To integrate and deploy them, a connector to Natoma’s MCP servers has also been announced.
Another major announcement, Snowflake is taking advantage of its global event to update CoCo (or Cortex Code), its data manipulation-oriented coding assistant. “It targets both business analysts and data engineers. One of its main strengths lies in its source-agnostic nature,” comments Bala Kasiviswanathan, vice president of AI & Developer Experiences at the publisher. Cortex CoCo notably supports AWS Glue, PostgreSQL without forgetting Databricks. During its summit, Snowflake unveils Cortex Coco Desktop. This tool comes in the form of a development environment that is installed directly on users’ workstations, whether they are on Windows or Mac.
Cortex Coco for developers
Here again, the approach is intended to be agnostic. Cortex Coco is now equipped with extensions for Visual Studio Code and Cursor, but also for Excel and especially for Claude Code, Anthropic’s model focused on coding. Added to this is a Slack bot as well as a mobile application to meet developers’ needs for collaboration. Cortex Coco also has interfaces to connect to the no-code development environments Superblocks and Retool. Downstream, Snowflake introduces cloud-based agents to autonomously execute tasks scheduled in Cortex Coco.
Ultimately, Snowflake draws the outlines of a big data platform in which AI is put at the service of data and not the other way around. This is the entire core of its value proposition.
