Snowflake CoCo Desktop became generally available, so I tried out all the new features and operational features

This is Kawabata.

Snowflake's AI development environment, CoCo Desktop, became GA (Generally Available) on July 21, 2026.

https://docs.snowflake.com/en/release-notes/2026/other/2026-07-21-coco-desktop-ga

This article covers the main new features since the Preview period (Memory, semantic code search, multiple sheets in SQL Playground, etc.) as well as the operational and governance features to be aware of at GA (usage history view, credit limits, Managed settings). In this article, the product name is written as "CoCo Desktop" in accordance with the current official notation.

Note that installation, basic operations, and MCP integration are covered in the following articles and will not be repeated here.

https://dev.classmethod.jp/articles/snowflake-cortex-code-desktop-install-basic-usage/

https://dev.classmethod.jp/articles/snowflake-cortex-code-desktop-mcp-mlit-real-estate/

Overview of CoCo Desktop GA

CoCo Desktop is a native desktop application for macOS / Windows, positioned as the standalone IDE version of the Cortex Code (CoCo) family.

The availability conditions and billing at GA are as follows.

• Available on Commercial accounts with cross-region inference enabled (Gov, VPS, and Sovereign accounts are not supported)
• When used with an existing Snowflake account, pay-as-you-go billing based on token consumption
• For individual developers, a subscription including a monthly usage allowance is available, with a 30-day free trial from registration. After the trial ends, it transitions to a paid subscription unless cancelled, and if the monthly usage allowance is exceeded, CoCo Desktop cannot be used until the next billing period
• Separate from CoCo Desktop usage, standard Snowflake charges apply for warehouses and storage used during SQL execution
• Snowflake's model terms apply to passthrough to the models used

Note: While CoCo Desktop itself is GA, within the Cortex Code family, MCP support, plugins, Agent SDK, and ACP support remain in Preview as of the time of writing. In particular, ACP is a feature for using CoCo CLI as an agent backend for editors / IDEs, and is not a feature of CoCo Desktop.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop

Prerequisites

• The active role requires the following database roles
  • SNOWFLAKE.COPILOT_USER
  • SNOWFLAKE.CORTEX_USER or SNOWFLAKE.CORTEX_AGENT_USER
• By default, SNOWFLAKE.COPILOT_USER and SNOWFLAKE.CORTEX_USER are granted to the PUBLIC role. SNOWFLAKE.CORTEX_AGENT_USER is not granted to PUBLIC by default and requires explicit granting to the role being used
• Cross-region inference must be enabled on the account

Note: Revoking SNOWFLAKE.CORTEX_USER from PUBLIC will remove access not only to CoCo Desktop but also to other Covered AI Features including Cortex AI Functions (AI_COMPLETE, etc.). If you want to control only CoCo Desktop usage, the official documentation recommends using the daily credit limit parameter described later.

• Semantic code search requires access to the snowflake-arctic-embed-l-v2.0 model

Verification Environment

• Verification date: July 22, 2026
• CoCo Desktop: v1.20.0
  
• OS: Windows
• Snowflake: Enterprise edition

Trying Out New Features

Memory (Remembering Across Conversations)

A feature that allows the agent to remember facts (preferences, project context, etc.) across conversations and automatically recall them in new conversations. The memory itself consists of Markdown files under ~/.snowflake/cortex/memory/, with two scopes: global memory shared across all workspaces, and project memory specific to a particular workspace.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/memory

It can be enabled and managed in the Memory section of Agent Settings → Personalization.

When you share a preference during a conversation, an "Updating memory" indicator appears and the information is written to memory.

When a new session is started, the content shared in the previous session was reflected along with the "Reading memory" display.

Note: Memory is for persistent context and does not save the temporary state of the current task. Also, Reset memories cannot be undone.

Custom instructions (Always-applied custom instructions)

Instructions that are always applied to the agent can now be managed on a dedicated page. Enter instructions in the editor at Agent Settings → Personalization and save them; they are saved to ~/.snowflake/cortex/AGENTS.md and applied to all workspaces.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/personalization

Configuration content

# Basic Policy
- Responses should be in Japanese
- Comments in code should also be written in Japanese
# SQL Conventions
- SQL keywords should be written in uppercase (SELECT, FROM, WHERE, etc.)
- Table and column names should be written in lowercase snake_case
- Do not use SELECT *; explicitly specify the required columns
# Snowflake-specific Rules
- Explicitly specify the WAREHOUSE for verification queries
- Warn before executing operations with cost impact (full scans of large tables, etc.)
# Output Format
- Summarize procedural explanations concisely in numbered lists

Let's check with Please tell me about the CoCo Desktop configuration procedure.

From the text in the red box, it appears that the Custom instructions settings have been applied.

On the same page, you can also control the on/off of built-in tools (Web Search Tool / Web Fetch Tool / Notebook Tools), and the toggle is applied immediately.

Note: The upper limit for Custom instructions is 512 KB. Also, using the Web Search Tool requires account-level permission (enabling the ENABLE_CORTEX_WEBSEARCH parameter).

Semantic code search (tgrep)

A feature that allows you to search a codebase by "meaning" using natural language. It uses Snowflake Cortex embeddings (snowflake-arctic-embed-l-v2.0) and can identify relevant code using queries like "the part that validates the connection before executing a query" rather than exact keyword matching.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/semantic-code-search

By default, indexing is not automatic and requires explicit opt-in by the user. Enable it from Agent Settings → Indexing.

1. Open Agent Settings → Indexing
2. Turn on Enable semantic search (tgrep) (a global setting shared across all workspaces)
3. Turn on Search this workspace (a setting per current workspace)
4. Build an index for the current workspace with Index workspace

※ The CoCo Desktop workspace contains jaffle-shop from dbt-labs.

https://github.com/dbt-labs/jaffle-shop

After building the index, when I asked the agent for the location of code in natural language, the relevant file was identified.

※ I found it difficult to clearly judge the effect of building the index with this small repository alone. On the other hand, since code and documentation are indexed in chunks, it seems particularly effective for larger repositories or cases where you want to search for specifications or implementation intent in natural language.

Note: Not available with remote / virtual workspaces such as Remote-SSH or dev containers (local execution only).

Context management and /compact

A context window that visualizes token consumption during a conversation, and a /compact command that summarizes the conversation and frees up tokens, have been added.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/context-management

Hovering over the context window in the lower right of the message input field displays a popover showing the breakdown of token usage.

Running /compact in the middle of a long conversation summarizes the conversation history and frees up tokens while continuing the same thread.

Change tracking panel (Changes)

When the agent modifies a file, a summary card showing "N files changed" appears, and you can review the changes using a unified diff, a changed file tree, and a scope selector. Accept / reject is also possible per file.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/agent-view-and-editor-view

Multiple sheets in SQL Playground

The lightweight SQL editor built into the right panel of Agent Manager now allows you to manage multiple SQL sheets (.sql buffers) per workspace.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/sql-playground

• How to open: SQL Playground icon in the right toolbar, "Open SQL Playground" in the Command Palette, or the "Open in SQL Playground" action from agent results
• Execute: Run button or ⌘Enter (macOS) / Ctrl+Enter (Windows). The statement at the cursor position is executed
• Sheet management: Create new, rename, delete, and organize by drag & drop in the sidebar

Can also switch to chart display

Note: SQL Playground sheets are not source files under the workspace; they are saved in a scratch area at ~/.snowflake/cortex/sql/<workspace-id>/. Sheets are auto-saved to the local area per workspace, so at least in the verification environment, sheets could still be used after restarting CoCo Desktop. If you want Git management or standard editor features, you can open them as editor tabs with "Open in Editor" (the sheet on the Playground side remains as is).

Trying Out Advanced Features

Since each of these could be the subject of a full article on its own, this article limits itself to minimal operation verification.

Agentic Notebook

A feature that allows the agent to directly operate Jupyter notebooks (.ipynb) from the chat. It autonomously performs cell creation, editing, execution, output reading, and DataFrame inspection (pandas / polars / PySpark / Snowpark supported). As a built-in feature, no additional feature enablement is required on the CoCo Desktop side. However, executing cells requires an available Jupyter kernel such as a local Python environment, conda environment, or existing Jupyter Server (if using Snowpark, dependencies such as snowflake-snowpark-python also need to be installed on the kernel side).

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/agentic-notebook

When I instructed the agent to create a Notebook that creates a sample DataFrame, cells were generated, and I could proceed while reviewing changes with Keep / Undo.

In sample_analysis.ipynb, create and execute cells that use Snowpark to aggregate total monthly sales from the SNOWFLAKE_SAMPLE_DATA.TPCH_SF1.ORDERS table and convert it to a pandas DataFrame.

Automations

Saved prompts that run on a schedule. A new chat session starts with each execution, and a desktop notification is delivered upon completion.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/automations

From the Automations view, create a new automation with New automation by setting the title, prompt, and frequency, and you can run it immediately with Run now.

I configured it to append to a log file. Settings such as execution frequency are available.

Let's try running it immediately with Run now.

Note: If execution reaches a tool call requiring approval, the Automation stops and waits for approval. If you want scheduled execution to complete unattended, it is recommended to run it once with Run now and approve the required tool execution permissions in advance.

Note: Automations run when CoCo Desktop is running and the PC is not in sleep mode. If the app is closed at the scheduled execution time, the run is skipped, but when the app is reopened, only the most recent single missed run per Automation is automatically executed as a catch-up run (older missed runs are discarded). Definitions are saved locally per machine/OS user and are not synced, and the execution history is a maximum of 50 entries.

Agentic Browser

The agent can operate the browser built into CoCo Desktop to browse the web and run tests for local apps. Chromium Developer Tools are also available for manual verification.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/agentic-browser

dbt integration

Automatically activated when dbt_project.yml is detected in the workspace, enabling lineage visualization, compiled SQL preview, data preview, and command bar from the DBT tab. There are two execution modes: Local mode (default) using the local dbt CLI, and Snowflake-managed mode.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/dbt-integration

dbt docs

Note: In Snowflake-managed mode, in addition to the Preview tab, the Command bar and Build / Compile / Show quick actions are also unavailable (dbt commands are run from the Run button and execution popover). While a local dbt CLI is no longer required, a profiles.yml directly under the project folder, an enabled Personal Database, and a valid Snowflake connection are still required. Also, the dbt project synced to the Snowflake workspace has a limit of a maximum of 20,000 files, including target/ and dbt_packages/.

Building apps

Supports two frameworks, Streamlit in Snowflake and Snowflake App Runtime, with the agent assisting from scaffolding through iterative development with a local development server to deployment on Snowflake. Deployment requires Snowflake CLI 3.19 or higher.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/building-apps

Clicking Build an app launches the agent as shown below, allowing you to develop apps in natural language.

Checking Operational and Governance Features

These are features to be aware of when considering production deployment at GA.

Checking usage history (CORTEX_CODE_DESKTOP_USAGE_HISTORY view)

A usage history view dedicated to CoCo Desktop is available in the SNOWFLAKE.ACCOUNT_USAGE schema. You can check credit usage for the past 365 days per user and request, and the TOKENS_GRANULAR / CREDITS_GRANULAR columns allow you to see a breakdown by model (input / cache_read_input / cache_write_input / output).

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/cortex-code-desktop-usage-history-view

-- Credit consumption by user for the past 30 days
SELECT USER_ID, USER_NAME, SUM(TOKEN_CREDITS) AS TOTAL_CREDITS
FROM SNOWFLAKE.ACCOUNT_USAGE.CORTEX_CODE_DESKTOP_USAGE_HISTORY
WHERE USAGE_TIME >= DATEADD('day', -30, CURRENT_TIMESTAMP())
GROUP BY USER_ID, USER_NAME
ORDER BY TOTAL_CREDITS DESC;

Note: This view only includes CoCo Desktop requests and does not include usage from CoCo CLI or CoCo in Snowsight.

Daily credit limit (CORTEX_CODE_DESKTOP_DAILY_EST_CREDIT_LIMIT_PER_USER)

A parameter that allows you to set an upper limit on daily credit consumption per user. The default is -1 (unlimited), and setting 0 completely blocks access. Setting a positive value blocks access when the estimated usage over a rolling 24 hours exceeds the threshold.

https://docs.snowflake.com/en/user-guide/cortex-code/credit-usage-limit

-- Check the current setting value
SHOW PARAMETERS LIKE 'CORTEX_CODE_DESKTOP%' IN ACCOUNT;

-- Set a limit at the user level (takes precedence over account level)
ALTER USER coco_test_user SET CORTEX_CODE_DESKTOP_DAILY_EST_CREDIT_LIMIT_PER_USER = 0;

After reaching the limit, an error indicating that the daily credit limit has been reached is displayed on CoCo Desktop, and the user cannot use it until the estimated usage over the rolling 24 hours falls below the threshold.

-- Remove the setting
ALTER USER coco_test_user UNSET CORTEX_CODE_DESKTOP_DAILY_EST_CREDIT_LIMIT_PER_USER;

Note: Setting 0 at the account level blocks all users. An operational pattern of "setting 0 at the account level and setting a positive value at the user level only for users who are permitted" is also possible.

Permission modes

The approval behavior during tool execution can be switched between two modes.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/permission-modes

Switch using the approval mode picker in the lower left of the chat input field. Note that this setting applies to the entire machine (all projects on the same machine).

Note: Even with Bypass Approvals, code execution in the browser (browser_evaluate / browser_run_code), clipboard reading (browser_read_clipboard), and launching sub-agents always require confirmation. Conversely, other operations, including destructive commands, are executed without confirmation, so please limit use to trusted inputs and workspaces.

Operational recommendation: Normally use Default Approvals as the standard, and limit Bypass Approvals to isolated verification workspaces or repetitive tasks whose content you fully understand. The official documentation also recommends using the agent terminal sandbox (Agent terminal sandbox) as an additional defense.

Managed settings (Centralized management for organizations)

A mechanism that allows administrators to enforce organizational policies via a JSON file. It is intended for distribution using device management tools such as Jamf or Intune.

https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code-desktop/managed-settings

For example, the following settings deny dangerous commands and remote skills, disable Bypass Approvals, prohibit persistence of conversation history, and display an organization banner.

{
  "version": "1.0",
  "permissions": {
    "deny": ["bash(rm:*)", "bash(curl:*)", "bash(wget:*)", "skill(remote:*)"],
    "defaultMode": "allow",
    "dangerouslyAllowAll": false
  },
  "settings": {
    "forceNoHistoryMode": true
  },
  "ui": {
    "showManagedBanner": true,
    "bannerText": "Managed by your organization.",
    "hideDangerousOptions": true
  }
}

Setting permissions.dangerouslyAllowAll to false blocks Bypass Approvals for the entire organization. When verified on an actual device, Bypass Approvals in the approval mode picker was grayed out, displaying "Disabled by admin policy" and making it unselectable.

Note: If the version field is missing or invalid, the app starts in fail-closed (restricted mode). Users cannot disable administrator settings.

Closing

In conjunction with the GA of CoCo Desktop, I verified the new features since the Preview period as well as the operational and governance features. In addition to improvements to the development experience such as Memory and semantic code search, I feel that the major point of GA is that the governance features needed for organizational deployment—usage history view, credit limits, and Managed settings—are now in place.

I hope this article is helpful in some way!