Decisions Developer MCP

Overview

Early Release Feature

Decisions Developer MCP is an Early Release feature. Customers should review and validate all AI-generated content, configurations, and platform changes before using them. AI-generated output may not always be accurate, complete, or appropriate for every scenario.

Decisions Developer MCP allows Users to connect supported third-party AI clients, such as Claude and Codex, to the Decisions platform and use them to assist with platform development and configuration.

The integration and available tools are considered stable; however, the guidance provided to AI clients continues to evolve. As a result, Users may notice differences in AI behavior and generated output between releases as the feature is refined.

Decisions Developer MCP provides a foundation for AI-assisted work in Decisions while maintaining customer responsibility for reviewing and validating all generated results before use.

Supported Capabilities

Note for Supported Capabilities

The capabilities listed below describe the actions available through Decisions Developer MCP tools. Actual AI-generated output may vary and should always be reviewed and validated before use.

Decisions Developer MCP can create and modify the following asset types:

Projects
Designer Folders
Public Folders
Data Types
Flows
Statement Rules
Truth Tables
Rule Tables
Forms
Scheduled Jobs
Project Documents
Custom Web Apps
Formulas
Inline Rules
Hidden Filter and Sort Rules
Default Forms
Case Instances (at runtime)
REST API Endpoints

Non-Supported Capabilities

Matrix Rules
Pages
Reports
Dashboards
Groups
Accounts
Standalone Formulas
Integrations
Connections

*Resources cannot be deleted through the available AI tools.

Considerations

AI-generated content, configurations, and platform changes should always be reviewed and validated before use.
The Decisions Developer MCP tools and integrations are considered stable and support AI-assisted development within the Decisions platform.
The guidance and instructions provided to connected AI clients are actively evolving and may change between releases.
As guidance evolves, Users may observe differences in AI-generated results, recommendations, and implementation approaches from one release to another.
Changes in AI behavior do not necessarily reflect changes in underlying platform capabilities.
Customers remain responsible for validating all generated content, configurations, and platform changes before deploying them to production environments.

Configuring the OAuth Application

Open the Edit OAuth Application dialog.
Enter a Display Name for the OAuth application.
Enable Public Client (PKCE only) by setting the toggle to True.
Copy the Client ID value and save it for use during client configuration.
Add the desired URLs in the Redirect URIs fields.
Once all fields have been completed, click Ok to save the information and exit the window.

Example

Create a new Project.
Navigate to Manage > Configuration > Settings, enable Allow Agents to Design with MCP, and click Create OAuth Connection.
Verify that the Connection Status shows Connected, select a Coding Agent, and click Save.
Users can click the link below, Advanced: Manage OAuth Application Manually, to edit existing OAuth applications.
Users can proceed to authenticate with their preferred coding agent. For this example, Claude Code was utilized.

Note for Utilizing Coding Agents

Coding agents can assist with creating and modifying supported Decisions assets; however, customers should review and validate all generated changes before implementation or deployment.