Zoho has released a new developer resource for teams building on Zoho CRM: a dedicated zoho-crm skill file, alongside an updated OpenAPI Specification (OAS) repository. Together, these are designed to make it easier for AI coding assistants such as Claude Code, Codex, Cursor and GitHub Copilot to work with Zoho CRM's APIs, functions, widgets and client scripts.

What has changed in the OAS repository

Following the launch of the V8 APIs, Zoho previously published OAS files in its public OAS Git repository. These files have now been converted from JSON to YAML, with the content refreshed to reflect the latest API changes and expanded descriptions intended to make the APIs easier to work with.

The repository has also been reorganised with AI tools and agents in mind. Each API resource now sits in its own folder, containing:

• A resource YAML file setting out the schema and properties for that resource.
• An index.md file giving an overview to help an AI agent understand the resource's context.
• An operations folder, split into subfolders for each individual API operation, each containing a compressed markdown summary (operation.md) plus request and response examples.

According to the announcement, this restructuring is meant to cut down the amount of information an AI agent has to process for any given operation, and Zoho has said new OAS files will be published weekly as further V8 API functionality is released.

What the zoho-crm skill does

The zoho-crm skill is described as an orchestration layer that sits across several Zoho CRM developer tools, choosing the right one for a given request:

• REST APIs – identifies the correct v8 API operation, works out the parameters and OAuth scopes from the OAS files, and either calls the API directly via MCP or builds a full cURL request.
• Deluge functions – generates and checks server-side Deluge scripts for CRM automation. At present this covers Deluge only, not Java, Node.js or Python functions.
• Client scripts – produces browser-side scripts using ZDK for things like record pages, list views, buttons, validation and dialogs.
• Widgets – builds embedded HTML/CSS/JS components using the CRM JS SDK and ZRC, including calls to CRM data and external APIs.
• COQL – builds and validates COQL queries for filtered, cross-module or aggregated CRM data, along with the correct endpoint and scopes.
• Saved queries – helps decide whether a Module, COQL, REST API, external database or cloud database source is the right fit, including variable tokens and associations.
• Connections – guides OAuth-based integrations with external services and how to reference them correctly from a widget, Deluge script or client script.

Installing and using the skill

Per Zoho's announcement, the skill is installed by running the following command in your preferred AI coding harness:

Once installed, developers can describe a CRM requirement in plain language, for example asking to fetch open deals for the current month, to conditionally show a field based on another field's value, or to create a follow-up task when a lead's score drops below a threshold. In GitHub Copilot within Visual Studio Code, the skill can be invoked directly using the /zoho-crm command, and it can also be wired into a broader agentic workflow.

How a request is processed

Zoho describes a five-step process behind the scenes:

1. Understand the goal. The skill works out the action, entity and context behind the request.
2. Select the right primitive. It decides whether REST API, Deluge, client script, widget, COQL, query or connection best fits the task.
3. Load the relevant references. It pulls in the appropriate documentation as the source of truth.
4. Generate the solution. It produces the code, query, API request or configuration guidance.
5. Validate the output. It checks the result against CRM syntax, parameters, scopes and known limitations before presenting it.

Points to keep in mind

Zoho is explicit that this skill is meant to speed up development rather than replace developer oversight. Its guidance includes:

• Review and test every generated solution before it goes into production.
• The first output may need refinement; providing feedback helps the skill improve its response.
• For anything specific to your own Zoho organisation's setup, include those details in the request.
• For integrations with other Zoho products or third-party tools, providing details and relevant documentation helps the skill produce a more accurate result.