Generate Agent JSON

The generate_agent_json tool creates an Agent definition (JSON) from a business objective that you describe. The tool does not create anything on your Pricefx partition. To change the definition, use the Update Agent JSON tool. To create the Agent on your partition, use the Deploy Agent tool.

Agent creation flow chart showing Generate Agent JSON, then optional Update Agent JSON, then Deploy Agent.
The Agent creation flow has three steps: the Generate Agent JSON tool creates the definition, the Update Agent JSON tool changes it (optional, repeatable), and the Deploy Agent tool creates the Agent on the partition.

Tool Name – generate_agent_json

Title – Generate Agent JSON

Type – Private

Access Profile – Operator

Permissions

The generate_agent_json tool reads the Agent configuration and the Datamart definitions of your Pricefx partition. The tool does not create or change anything on the partition. The tool is available to the Operator access profile, so every user on a private connection can use it. The tool is available only when the agent generator is set up on the MCP server.

Parameters

Parameter

Required

Description

business_objective

Yes

What the Agent should do, in business terms. Name the outcome to watch and the data it concerns, for example Alert me when the margin of a product drops more than 5% month over month.

datamart_unique_names

No

Technical names of the Datamarts the Agent should analyze, for example Standard_Sales_Data. The default is the Datamart set in the partition's Agent configuration. Get the names from the List Data Sources tool.

Output

Field

Description

objective

Business objective the Agent was generated for.

agent

Generated Agent definition. Pass it to the Update Agent JSON or Deploy Agent tool without changes.

datamarts_used

Technical names of the Datamarts the generated Agent reads from. This list can differ from datamart_unique_names, because the generator can choose other Datamarts. explanation then says why.

explanation

Notes from the generator about how it built the Agent. Can be empty.

validation_status

Validation status that the generator reports.

validation_errors

Problems that the generator found in the Agent definition. Empty when there are none.

Behavior and Limitations

  • The tool generates one Agent definition per call.

  • Generation usually takes 30 – 120 seconds, because the generator uses an AI model. Many AI clients stop waiting after 60 seconds. If the generation times out, increase the tool timeout of your AI client, for example to 180 seconds, and try again.

  • The tool returns the Agent definition even when the generator reports validation errors. Check validation_errors before you deploy the Agent.

  • Each call generates a new definition. The same business objective can produce a different Agent each time.

  • The generated definition is not saved anywhere. If you start a new conversation, generate it again.

Errors

Situation

Error Message

A name in datamart_unique_names is not a Datamart on the partition

Unknown datamart uniqueName(s): [...]. Available datamarts: [...]

The generator returns no Agent

Agent generation returned no agents: …

The generator fails to create the Agent

Agent generation failed: …

The generation takes longer than your AI client waits

A timeout error from your AI client

See Also