Palantir AIP

Palantir AIP (Artificial Intelligence Platform) is Palantir's enterprise-grade layer for building, deploying, and governing LLM-driven applications and agents on top of Foundry and Gotham. It exposes a company's data, logic, and actions as an 'Ontology' that language models can query, reason over, and act on — with human-in-the-loop approvals, permissioning, and audit trails baked in. AIP Logic lets teams compose multi-step agentic workflows, AIP Assist is a copilot inside the platform, and AIP Agents run autonomously against defined objectives.

Where general AI platforms hand you a model and a chat box, AIP's differentiator is the Ontology and its 'kinetic' integration with operational systems — the same model can read a work order, propose a schedule change, and actually execute it against SAP or a warehouse-management system, subject to policy. It's aimed squarely at defense, intelligence, industrials, healthcare, and Fortune 500 operations teams that already run Foundry; pricing is enterprise-only, quoted via sales, and typically bundled with a Foundry engagement. There is no self-serve free tier, though Palantir runs frequent AIP bootcamps for prospective customers.

AIP is model-agnostic: it routes to GPT-4-class models, Claude, open-source Llama variants, and customer-hosted models, and supports air-gapped and classified deployments including IL5/IL6. Integrations are less about SaaS connectors and more about the deep data-pipeline and action-graph tooling Palantir has built over two decades. The caveat is scope: this is not a tool you spin up on a credit card, and the value only appears once your data and processes are modeled in the Ontology.