AI & agents
Logic Conductor
Turn the logic blocks, logic flows and business processes you choose into tools that AI assistants and agents can run.
Applies toDataverse
Logic Conductor (long form: Logic Fabric Conductor) is an MCP server that offers your finished business logic to AI. Every logic block, logic flow and business process you mark with Expose as MCP Tool becomes a tool that an AI assistant or agent can call, with the inputs, outputs, title and description you defined. Instead of teaching an AI how to calculate a discount or approve a credit limit, you let it call the rule your business already runs on.
Where Logic Foundry lets an AI build and change your logic, Logic Conductor lets an AI run it.
What becomes a tool
Section titled “What becomes a tool”| You expose | The tool | What a call does |
|---|---|---|
| A logic block (formula, validation, validation set, decision table, decision tree) | One tool named after the block | Calculates the outputs from the inputs. Doesn’t change data. |
| A logic flow | One tool named after the flow | Runs the flow and returns its outputs and whether it completed. Can change data. |
| A business process | start_<process name> |
Starts the process on an existing record and returns the new process instance’s ID and subject. Requires Process Orchestrator. |
Only artifacts with Expose as MCP Tool turned on, in the projects your administrator chose, become tools. Everything else stays invisible to AI.
What the AI sees
Section titled “What the AI sees”- A tool name made from the artifact name, in lowercase with spaces replaced by underscores, for example
approve_credit_limit. When Conductor exposes more than one project, the project name is added in front:sales_approve_credit_limit. - A description taken from the artifact’s description. The AI reads it to decide when to call the tool.
- Typed inputs from the artifact’s input arguments, each with its own description.
- Typed outputs from the artifact’s output arguments. Lists come back as typed lists and table records as structured records.
Example
Section titled “Example”Your sales team asks their AI assistant: “Can I give Contoso a 15% discount on this quote?”
The Sales project has a decision table Max Discount exposed as an MCP tool, with inputs Customer Tier and Quote Total and an output Max Discount %. The assistant reads the account’s tier, calls max_discount, gets 10, and answers: “No. Contoso’s tier allows up to 10% on a quote of this size.” The answer comes from your rule, not from the AI’s guess, and it runs with the user’s own Dataverse permissions.