Snowflake CoCo Goes Everywhere

Snowflake is pushing its AI coding agent, CoCo, beyond its data cloud to meet developers where they work. Announced at Snowflake Summit 2026, CoCo is rolling out a native desktop application for Windows and macOS, alongside cloud agents, an SDK, and upcoming mobile and Slack integrations.

StartupHub data

Companies working on this

Profiles of the companies named in this story, with funding and a one-liner from our database.

Anthropic

Private / $100B+ est

Anthropic is an AI safety and research company building reliable, interpretable, and steerable AI systems, best known for the Claude family of models.

OpenAI

Private / $100B+ est

OpenAI is an AI research and deployment company dedicated to ensuring that artificial general intelligence benefits all of humanity.

Thomson Reuters

Global provider of business information, news, and technology for legal, tax, audit, and accounting professionals.

Snowflake

$12.4B

A cloud-based data platform enabling data warehousing, data lakes, data engineering, and data sharing.

The move signals a shift from theoretical AI potential to practical, governed enterprise deployment. While early AI assistants offered basic code completion, the next wave, including CoCo, is agentic, capable of inspecting codebases, reasoning through tasks, and managing workflows with human oversight.

Generic AI coding agents often falter when faced with the complexities of enterprise data environments, lacking the necessary grounding in data context and permissions. CoCo aims to close this gap, acting as a control plane for developers working with modern data stacks.

Thomson Reuters is already leveraging CoCo to accelerate modernization and AI pipeline scaling, delivering insights in days rather than weeks within a governed environment.

CoCo's Performance Edge

Snowflake claims CoCo sets a new bar for data engineering AI. On the ADE-Bench framework, CoCo achieved a 72.1% pass rate, surpassing Anthropic's Claude Code and OpenAI's Codex (both at 65.1%).

Critically, CoCo achieves this lead with greater efficiency, using 51% fewer tokens and taking 8% less time than Claude Code. This is attributed to targeted exploration of relevant data and a native tools approach, leveraging Snowflake, dbt, and Airflow directly.

Expanding the Agentic Ecosystem

The expansion transforms CoCo into a full AI development platform. Cloud Agents are the foundation, enabling long-running workflows across various surfaces, including direct integration within Snowsight for browser-based CLI power without local setup.

CoCo Desktop offers a unified, governed environment for building data pipelines, applications, and debugging notebooks, aiming to keep developers in flow from prototype to production.

Automations allow for scheduled, autonomous workflows, such as pipeline refreshes and data quality checks, all governed by Snowflake's RBAC and audit trails.

Extensibility through MCP integrations and a growing catalog of skills will allow teams to connect CoCo to existing systems and codify best practices.

Future Integrations

Upcoming Slack and mobile app integrations will allow users to interact with CoCo directly within chat or on the go. The Slackbot will provide a governed interface for data work, grounded in enterprise data and user permissions.

The mobile app targets leaders and managers needing to monitor pipelines, review outputs, and approve workflows remotely.

A Platform for Developers

CoCo is also positioned as a platform for developers to build upon. The CoCo Agent SDK provides programmatic access to the agent's capabilities, allowing custom applications and domain-specific workflows.

This SDK enables developers to embed data-native agentic capabilities directly into their stacks, from pipeline scripts to backend services.

Governance Remains Paramount

Snowflake emphasizes that CoCo is built with AI-native development and enterprise-grade security. All operations adhere to Snowflake's RBAC, keeping LLM inference within its security perimeter and protecting against risks like prompt injection.

Layered guardrails, prompt logging, query tagging, and usage monitoring ensure auditability and governance for safe, scaled deployment.