In this post, we show how the Model Context Protocol (MCP) enables seamless integration between voice agents and Palantir’s ontology.

Many of our clients have built rich ontologies in Palantir’s AIP platform to power key workflows and automations, but call centres have remained difficult to automate without modern Speech to Text (STT), Text to Speech (TTS) and reasoning LLMs.

By taking advantage of AIP’s flexible Ontology SDK (OSDK), we were able to build an MCP server to connected business logic captured by the ontology with advanced AI voice platforms such as ElevenLabs, enabling intelligent, voice-driven workflows.

Components:

1. AIP Ontology: Data model and Business Logic implemented in Palantir AIP that securely connects to enterprise data systems. This can be exposed to external applications using OSDK.
2. Node JS MCP Server: Typescript application that defines MCP tools that utilize OSDK to retrieve data, run actions, and write data. This is where we provide detailed prompt templates that help LLMs gain context on the structure and functionality of our ontology.
3. Voice to Voice Client Application: React web app built using ElevenLabs Client SDK, that connects to its Conversational AI API and interacts with our MCP Server to gain context and run MCP tools.

Case Study

A multinational property management company we partnered with struggled to streamline its operations. Managers and brokers across multiple locations spent excessive time balancing office tasks such as lead management, customer relations, and inventory tracking with on-site responsibilities like property showings, lease negotiations, and finalizing deal paperwork.

Key data on inventory, pricing, promotions, and financing was locked in desktop-only proprietary apps, making it inaccessible in the field where it’s needed most. This dependence on back-office support slowed their customer communication and delayed deal closures.

Using our voice-to-voice Sales Assist Agent, running on a smartphone browser, their employees are now able to get realtime back office support by sharing compound question & task list like the following:

Our customer Atlas Solutions Inc. is looking to expand to a second office that has a 100 person capacity, they would like to be in close proximity to public transportation and their budget is similar to their current space. Additionally, they would like to be close to their main client Beacon Inc.’s main campus. Could you suggest inventory that satisfies this need and set up meetings with property managers in the top three locations for Tuesday morning.