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

01 AI Can Write Code, but It Still Doesn’t Know How Enterprises Operate

Coding agents can generate functions, modify files and even complete an application, but the real challenge in enterprise settings is ensuring that generated code acts correctly on real business objects. Traditional systems scatter business knowledge across databases, API docs, code and human expertise, making it hard for agents to verify whether an action is permissible.

02 SuperRepo Is More Than a Code Repository – It Unifies the Business World

SuperRepo, announced as a beta on 2026‑08‑04, is defined as a professional monorepo that stores Ontology definitions, Functions and React applications together. In conventional projects these three layers live in separate tools and repositories, leading to semantic drift when business objects change. SuperRepo collapses this into a single edit‑preview‑build cycle, with the Foundry CLI launching local services that embed Ontology, route Functions‑backed Actions, and connect OSDK‑based apps.

03 Why “Ontology‑as‑Code” Beats Adding a Semantic Layer

Semantic layers answer “what does this field mean”. Ontology‑as‑Code also answers the current state of an object, its relationships, which actions are allowed, and which Functions implement those actions. By bringing these definitions into code, they become subject to branch development, code review, automated testing, CI/CD and rollback, turning business semantics into a first‑class engineering artifact.

04 Core Capabilities of SuperRepo

Business Model : TypeScript Ontology‑as‑Code and import of existing Ontology resources.

Functions : Currently TypeScript v2 Functions; Python Functions are planned.

Applications : React/OSDK apps with local preview support.

Agent Integration : Accessible via Ontology MCP or Palantir MCP; respects Foundry OAuth 2.0 permissions.

Data & Compute : Objects can declare data sources; PySpark pipelines and Compute Modules are not yet supported.

Delivery : Produces reproducible Marketplace products deployable via internal or external CI.

05 Testing, Linting and Governance

Embedded Ontology enables cross‑component integration tests that cover object definitions, Functions and application flows. Ontology owners can author lint rules to enforce naming conventions, structural patterns and design guidelines before deployment.

06 Beta Limitations

SuperRepo is still beta: not all Foundry environments expose it, and capabilities may change. Functions are limited to TypeScript v2; Python Functions, external data‑source calls, Agent Engine, Automate and full data pipelines are still in development. Direct definition of PySpark transforms and Compute Modules is unavailable, and the repository defaults to a locked state that can be overwritten on the next deployment.

07 The Strategic Goal: A Governed Business Runtime for AI Agents

Palantir’s long‑standing moat has been data integration and large‑scale deployments. SuperRepo extends this by productising the entire stack—data, business semantics, Functions, Actions, permissions and applications—into a single development and delivery pipeline. By making the business world itself codified, AI agents receive a governed, versioned runtime rather than a set of ad‑hoc APIs, reducing the risk of mismatched models, stale logic and unauthorized actions.

In summary, Ontology‑as‑Code attempts to bring business semantics back into the software engineering lifecycle, allowing developers and AI agents to build, test and deploy enterprise applications on a shared, versioned business type system. The approach is still early, but its direction is clear.