Connector between Claude and CRM from Dynamics Microsoft

Hi Javier, should the final production architecture keep Python as the main natural-language connector with Node/Express calling into it, or do you want the full Claude tool-use and Dataverse operation logic rebuilt in Node.js with feature parity? Also, what would you consider the acceptance test for completion: successful sandbox creation of quotes, quote lines, products, accounts and leads, including price list/unit-of-measure cases, or validation against the live environment after sandbox testing?

Could you share the current architecture of the prototype (how Python and the JavaScript data layer interact with Dynamics 365/Dataverse), and specify which parts are already stable versus which areas are failing or still experimental?

A few quick questions:
1. Is the current Python prototype fully working in your sandbox environment?
2. Do you already have all Claude prompts and tool schemas finalized?
3. For the Node.js version, do you want an exact copy of the Python connector or are architecture improvements also welcome?

Thanks for sharing the brief. A few questions that would help me accurately scope the work:

• Is the existing Python prototype fully functional against a sandbox Dataverse environment, or has it only been tested locally?

• Approximately what percentage of the required functionality is already working today versus what still needs development?

• Do you already have a preferred deployment environment for the Node.js connector (on-premise server, Azure, AWS, Docker, etc.)?

• Will the Node.js version completely replace the Python connector, or will both need to be maintained going forward?

• Are there any custom Dynamics entities, plugins, or business rules that the connector must respect in addition to the standard Accounts, Leads, Products, Quotes, and Quote Lines?

• Have the supplier/customer classification fields already been identified, or is metadata discovery still required?

• For quote creation, are price lists and units of measure standardized across the environment, or do they vary by business unit?

• Do you expect bulk operations involving hundreds or thousands of records regularly, or is the batch functionality mainly for future scalability?

• Should all user actions be logged for auditing purposes, even if the optional audit-trail feature is deferred?

• Will staff interact with the connector through a command-line workflow initially, or is there already an existing web interface where the Node.js version will be embedded?

• Are there any security, compliance, or approval requirements that affect how prompts, responses, and created records should be stored?

• What is the target timeline for production deployment once sandbox validation is completed?