Prophecy Agent overview

The Prophecy Agent uses natural language to build, modify, and interpret pipelines, analyses, documentation, and related project artifacts. When you describe your intent, the Agent inspects your active project, searches metadata, retrieves sample data where permitted, and generates or updates transformations directly within the project.

​

Why use the Agent?

The Agent accelerates pipeline development by reducing the manual steps required to build and modify data workflows. Instead of navigating multiple interfaces, you describe intent. The Agent translates that intent into concrete, inspectable project changes within your active environment. The Agent assists development; it does not autonomously deploy or promote changes.

​

Scope and control

The Agent operates within the currently active project and under the permissions of the user who invoked it.

• All SQL execution and pipeline runs occur under your existing credentials.
• All changes remain visible, versioned, and editable within Prophecy.

Prophecy includes specialized agents for transformation, harmonization, and documentation workflows. The Transform Agent runs separately from the Documentation and Harmonization Agents. To use the Harmonization Agent or Documentation Agent, you must disable the Transform Agent. (Harmonization and Documentation can be used without disabling one another.)

• Transform Agent (default) — Builds and modifies pipelines, analyses, and documentation using natural language.
• Harmonization Agent — Automates mapping source data to a defined Common Data Model (CDM).
• Documentation Agent — Generates complete project and pipeline documentation.

​

Workflow

Agent workflows vary depending on the active agent.

​

Transform Agent workflow (default)

The Transform Agent focuses on building and modifying pipelines, analyses, and related project artifacts within the active project. When you submit a request, the Agent:

• Interprets your intent.
• Inspects the current project graph and schema metadata.
• Generates or modifies pipeline logic.
• Validates transformations and project structure.
• Surfaces changes for review.

The Transform Agent supports two primary intents throughout the pipeline development lifecycle:

• Exploration for discovering and understanding data sources.
• Transformation for building and modifying pipeline logic.

Use the Agent to search your data warehouse, preview datasets, and validate data quality before building transformations. During transformation, describe data operations in natural language to generate gems and modify pipeline logic.

​

Harmonization workflow

The Harmonization Agent focuses on mapping source schemas to a defined Common Data Model (CDM). The workflow typically includes:

• Defining or selecting a CDM.
• Generating source-to-target mappings.
• Reviewing confidence indicators and data quality tests.

​

Agent architecture

The Prophecy Agent operates within Prophecy’s structured execution environment and is equipped with tools to query warehouses, compile code, run pipelines, and modify project artifacts directly. Rather than generating text alone, the Agent uses controlled internal tools to inspect, modify, and validate project assets directly within the active project. This enables the Agent to reason over:

• Your project’s structure (pipelines, analyses, datasets, documentation).
• Schema metadata and dataset relationships.
• Version-controlled project artifacts.

​

Tool-based execution

The Agent uses a set of internal tools to take concrete actions inside your project, including:

• Querying your connected warehouse.
• Retrieving metadata and sample data.
• Compiling and validating code.
• Running pipelines.
• Creating and editing project artifacts.

When the Agent executes SQL or runs a pipeline, it does so under the permissions of the user who invoked it. All warehouse interactions respect existing access controls and governance policies. Before applying changes, the Agent compiles and validates generated logic against your project structure to detect syntax errors, broken references, or incompatible transformations.

​

Schema and lineage awareness

The Agent understands your project’s schema metadata and dataset relationships. When generating or modifying transformations, it:

• Inspects upstream schema definitions.
• Validates generated changes against your project structure before applying them.
• Surfaces potential conflicts or mismatches when detected.

The Agent assists with structural consistency but does not override warehouse-level schema constraints.

​

Project boundary

All Agent activity is strictly contained within the active project and cannot affect external projects or platform-level configuration. It can:

• Create and refactor pipelines.
• Generate and update analyses.
• Modify documentation.
• Edit datasets and other project artifacts.

​

Single-agent editing

Only one Agent session can edit a project at a time. Collaborative agent editing across multiple users is not supported.

​

Human in the loop

The Agent generates and can apply changes inside your project, but you remain in control. All modifications are visible in the editor and can be inspected, edited, or reverted. You can:

• Inspect every transformation.
• Validate data samples.
• Modify generated logic.
• Re-run pipelines as needed.

The Agent does not replace review. Users remain responsible for validating logic, data correctness, and production readiness.

​

Available features

​

Features in progress

Features in progress are subject to the same project-scope and permission constraints as existing capabilities.