--- title: Snowflake's Cortex and CoCo updates bring cost controls,... description: Snowflake has shipped a range of updates across its CoCo and Cortex Agents platforms. CoCo Desktop is now generally available on macOS and Windows, offering... canonical: https://daily.dev/posts/snowflake-s-cortex-and-coco-updates-bring-cost-controls-native-context-and-async-execution-to-ente-yoclps0zu twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Snowflake's Cortex and CoCo updates bring cost controls, native context, and async execution to enterprise AI | daily.dev og:description: Snowflake has shipped a range of updates across its CoCo and Cortex Agents platforms. CoCo Desktop is now generally available on macOS and Windows, offering... og:url: https://daily.dev/posts/snowflake-s-cortex-and-coco-updates-bring-cost-controls-native-context-and-async-execution-to-ente-yoclps0zu og:image: https://api.daily.dev/og/posts/YOcLpS0zU.png og:image:alt: Snowflake's Cortex and CoCo updates bring cost controls, native context, and async execution to enterprise AI og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Snowflake's Cortex and CoCo updates bring cost controls, native context, and async execution to enterprise AI **[Collections](https://daily.dev/sources/collections)** · 4 min read · 3 upvotes · 0 comments ## Summary Snowflake has shipped a range of updates across its CoCo and Cortex Agents platforms. CoCo Desktop is now generally available on macOS and Windows, offering native access to local files and Snowflake context including schema, RBAC policies, and masking rules. Cloud Agents are also GA in Snowsight. On the cost side, Snowflake added per-user spending quotas, real-time cost visibility, and reports roughly 20-28% cost reductions. Cortex Agents gains Skills Packages, Tool Search for dynamic runtime discovery, an Async Agent API for long-running tasks, a sandboxed Python 3.12 Code Execution Tool, Interrupt and Resume for human-in-the-loop workflows, Partial Access for serving users with different permission levels via existing RBAC, and a Versioning UI for agent configuration management. ## Content Snowflake has been shipping a lot of updates to its CoCo (Cortex Coding) assistant and the broader Cortex Agents platform. Here's a consolidated look at what's changed, what's new, and how to actually use it. ## CoCo is available in more places now CoCo Desktop is generally available for macOS and Windows. It's a native app with access to local files, repositories, and your Snowflake context baked in. Cloud Agents are also generally available in Snowsight, so you can run web search, shell commands, and Python scripts directly in the browser without installing anything locally. CoCo Mobile for iOS and Android is in private preview if you want to try it early. For teams that need remote access, the CLI has an experimental server mode. Set the `CORTEX_CODE_EXPERIMENTAL_FEATURES` environment variable with `webServer:true`, then run `cortex --ui`. It serves on `localhost:4242` by default, with configurable host, port, and allowed origins. The browser UI supports file and image management, multi-session chat, Agent Teams monitoring, skill browsing, Git worktree management, and scheduled task configuration — useful for long-running workflows or development on remote machines. ## Why platform-native matters for data work Generic coding agents produce syntactically valid code that's often contextually wrong for data work. They don't know your schema, your RBAC policies, your masking rules, or your row access policies. You end up manually injecting context, and that becomes a maintenance burden fast. CoCo runs under your actual Snowflake role, reads your catalog directly, and ships with built-in skills for Snowflake-specific tasks: `ACCOUNT_USAGE` queries, lineage tracing, PII classification. The scaffolding you'd otherwise build yourself is just there. ## New Cortex Agents capabilities Beyond CoCo itself, Snowflake is expanding the Cortex Agents platform: - **Coding Agent** backed by the CoCo runtime - **Skills Packages** for shared capability layers across agents - **Agent Toolsets** for reusing tool definitions - **Tool Search** for dynamic tool discovery at runtime - **Async Agent API** for long-running background tasks - **Code Execution Tool** — sandboxed Python 3.12 with numpy and pandas, running inside Snowflake - **Interrupt and Resume** for human-in-the-loop steering mid-task - **Partial Access** so a single agent can serve users with different permission levels using existing RBAC - **Versioning UI** with visual configuration history, rollback, and promotion workflows Most of these are in public preview or generally available soon. ## Controlling AI spend Agentic sessions consume credits based on token processing, and costs can escalate quickly if you're not watching. Snowflake has added several controls. Snowflake reports about a 28% reduction in cost per prompt from internal efficiency improvements, with roughly 20% average cost reduction across customer usage. That helps, but you still want guardrails. Here are seven concrete controls: **1. ACCOUNT_USAGE history views** Query usage across CLI, Desktop, and Snowsight surfaces for full visibility. **2. Built-in /cost-intelligence skill** Ask CoCo natural-language questions about your spend directly. **3. Per-surface daily credit limits** Set via account parameters with user-level overrides. ```sql ALTER ACCOUNT SET CORTEX_COCO_DAILY_CREDIT_LIMIT = 100; ``` **4. Per-user quotas** Now a first-class Snowflake object in public preview. Enforces hard blocks across AI domains including Cortex Agents and CoWork. ```sql CREATE USER QUOTA my_quota CORTEX_CREDITS_QUOTA = 50 RESET_INTERVAL = DAILY; ``` **5. Forecast-based budgets** Early warnings with configurable notification channels before you hit limits. **6. Model access control via RBAC** The old `CORTEX_MODELS_ALLOWLIST` is deprecated and being retired by November 2026. Use role-based access controls instead. **7. Automated guardrails** Snowflake alerts and tasks can cancel runaway queries automatically. ## Admin controls Agent Settings (public preview) lets admins configure default models and push configurations across teams via role-based access controls. Per-user quotas work the same way — set them once, enforce them everywhere. The combination of cost visibility, hard quotas, and RBAC model controls gives you the levers to run AI at scale without surprise bills. ## Questions this post answers ### What is CORTEX_MODELS_ALLOWLIST deprecation timeline in Snowflake and what should I migrate to? CORTEX_MODELS_ALLOWLIST is deprecated and being retired by November 2026. The replacement is RBAC-based model access control, where you grant usage on specific Cortex models to roles directly: GRANT USAGE ON CORTEX MODEL snowflake-arctic TO ROLE analyst_role. This gives finer-grained, role-level control over which AI models each team can access. _Teams migrating Snowflake AI governance before the November 2026 cutoff track deprecation timelines like this on daily.dev._ ### How do I set per-user credit quotas for Snowflake Cortex AI usage? Per-user quotas are now a first-class Snowflake object in public preview. You create them with CREATE CORTEX BUDGET, specifying the user, a credit limit, and a period (e.g., DAY). They enforce hard blocks across AI domains including Cortex Agents and CoWork. You can also set account-level limits with ALTER ACCOUNT SET CORTEX_CREDITS_QUOTA and user-level overrides with ALTER USER SET CORTEX_CREDITS_QUOTA. _Engineers rolling out Snowflake AI tooling to broader teams find spend governance patterns like these on daily.dev._ ### What sandboxed code execution environment does Snowflake Cortex Agents use? Snowflake Cortex Agents includes a Code Execution Tool that runs sandboxed Python 3.12 with numpy and pandas, executing entirely inside Snowflake. It is part of the expanded Cortex Agents platform alongside capabilities like an Async Agent API for long-running background tasks and Interrupt and Resume for human-in-the-loop agentic workflows. _Developers building data agents on Snowflake keep up with platform capability changes on daily.dev._ ## Similar posts on daily.dev - [Governed AI for Every Builder: Enterprise Controls in Snowflake CoCo](https://daily.dev/posts/governed-ai-for-every-builder-enterprise-controls-in-snowflake-coco-uk5mcnnud) · Snowflake · 0 upvotes · 0 comments - [Cortex Code Updates: Faster AI Data Engineering on Snowflake](https://daily.dev/posts/cortex-code-updates-faster-ai-data-engineering-on-snowflake-usty05fvq) · Snowflake · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#snowflake](https://daily.dev/tags/snowflake) [View this post on daily.dev](https://daily.dev/posts/snowflake-s-cortex-and-coco-updates-bring-cost-controls-native-context-and-async-execution-to-ente-yoclps0zu) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Snowflake's Cortex and CoCo updates bring cost controls, native context, and async execution to enterprise AI","url":"https://daily.dev/posts/snowflake-s-cortex-and-coco-updates-bring-cost-controls-native-context-and-async-execution-to-ente-yoclps0zu","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/snowflake-s-cortex-and-coco-updates-bring-cost-controls-native-context-and-async-execution-to-ente-yoclps0zu"},"datePublished":"2026-08-04T21:03:45.270Z","dateModified":"2026-08-07T14:49:45.888Z","description":"Snowflake has shipped a range of updates across its CoCo and Cortex Agents platforms. CoCo Desktop is now generally available on macOS and Windows, offering...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/76e2796a8e98924b723004e35942a059?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/76e2796a8e98924b723004e35942a059?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/snowflake-s-cortex-and-coco-updates-bring-cost-controls-native-context-and-async-execution-to-ente-yoclps0zu","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,snowflake","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"Snowflake's Cortex and CoCo updates bring cost controls, native context, and async execution to enterprise AI"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/snowflake-s-cortex-and-coco-updates-bring-cost-controls-native-context-and-async-execution-to-ente-yoclps0zu#faq","mainEntity":[{"@type":"Question","name":"What is CORTEX_MODELS_ALLOWLIST deprecation timeline in Snowflake and what should I migrate to?","acceptedAnswer":{"@type":"Answer","text":"CORTEX_MODELS_ALLOWLIST is deprecated and being retired by November 2026. The replacement is RBAC-based model access control, where you grant usage on specific Cortex models to roles directly: GRANT USAGE ON CORTEX MODEL snowflake-arctic TO ROLE analyst_role. This gives finer-grained, role-level control over which AI models each team can access. Teams migrating Snowflake AI governance before the November 2026 cutoff track deprecation timelines like this on daily.dev."}},{"@type":"Question","name":"How do I set per-user credit quotas for Snowflake Cortex AI usage?","acceptedAnswer":{"@type":"Answer","text":"Per-user quotas are now a first-class Snowflake object in public preview. You create them with CREATE CORTEX BUDGET, specifying the user, a credit limit, and a period (e.g., DAY). They enforce hard blocks across AI domains including Cortex Agents and CoWork. You can also set account-level limits with ALTER ACCOUNT SET CORTEX_CREDITS_QUOTA and user-level overrides with ALTER USER SET CORTEX_CREDITS_QUOTA. Engineers rolling out Snowflake AI tooling to broader teams find spend governance patterns like these on daily.dev."}},{"@type":"Question","name":"What sandboxed code execution environment does Snowflake Cortex Agents use?","acceptedAnswer":{"@type":"Answer","text":"Snowflake Cortex Agents includes a Code Execution Tool that runs sandboxed Python 3.12 with numpy and pandas, executing entirely inside Snowflake. It is part of the expanded Cortex Agents platform alongside capabilities like an Async Agent API for long-running background tasks and Interrupt and Resume for human-in-the-loop agentic workflows. Developers building data agents on Snowflake keep up with platform capability changes on daily.dev."}}]} ```