How Palantir’s SuperRepo Turns Ontology into Code for Enterprise AI

Palantir recently announced SuperRepo, a monorepo‑style development environment that goes beyond merely hosting code. It allows developers to define business objects, relationships, interfaces and actions using TypeScript, and to place these Ontology definitions alongside Functions and React applications in the same repository.

Ontology Moves into the Software Lifecycle

Historically, Ontology answered the question “what composes the enterprise business world.” SuperRepo pushes this further to ask “how to continuously develop and deliver applications around this business world.” By integrating Ontology into local preview, unified build and versioned artifacts, the business model becomes part of the standard code review, automated testing, CI/CD and rollback processes.

Key Capabilities of SuperRepo

TypeScript‑based Ontology‑as‑Code: developers declare object types, relationships, Actions and Functions directly in code.

Automatic OSDK generation: changes in Ontology trigger regeneration of type bindings that both Functions and front‑end applications can consume.

Local embedded Ontology: the Foundry CLI launches services that emulate the real Foundry environment, enabling end‑to‑end preview of Ontology, Functions and applications.

Single reproducible product: the whole stack is packaged as a Marketplace product and can be deployed via Foundry CI or external CI systems.

Integration with Palantir MCP

SuperRepo is positioned alongside Palantir’s Model‑Centric Platform (MCP). While MCP focuses on platform construction and application development, Ontology MCP exposes the defined objects, Actions and query Functions to external agents under strict OAuth 2.0 authentication and permission checks. This separation ensures that agents receive a constrained, policy‑enforced view of the business model.

Why “Ontology‑as‑Code” Matters

Traditional semantic layers only describe the meaning of metrics or fields. Ontology‑as‑Code also captures the current state of objects, their relationships, permissible actions and the Functions that implement those actions. By treating these definitions as code, they inherit versioning, testing and deployment discipline, reducing the risk of semantic drift across data, API and UI layers.

Current Limitations

SuperRepo is still in beta. Not all Foundry environments support it, and several capabilities remain in development: Python Functions, external data source calls, full Agent Engine integration, PySpark transforms and Compute Modules are not yet available. The product may change before GA.

Strategic Implications

Palantir’s long‑term goal appears to be owning the business‑runtime layer for enterprise AI. By unifying data, semantics, Functions, Actions, permissions and applications in a single release unit, the company creates a sticky platform that goes beyond data integration, making it harder for customers to migrate away without rebuilding their entire business model.

In summary, SuperRepo demonstrates a concrete step toward “Ontology‑as‑Code,” where business semantics are managed with the same rigor as application code, providing a versioned, testable, and auditable foundation for AI agents operating within enterprise environments.