The TARS Model Context Protocol Server is designed to work with Claude or ChatGPT to supply the Large Language Model with internal company and context specific Information. Through a standardised interface the Model can query data provided to find the relevant information quickly, use it in inference and provide the original documents for the user to check the model reasoning.
Data for the MCP Server can be completely unstructured and available in a number of formats including all MS Office formats, PDF, images, call transcripts etc. Overnight any new information made available will be transformed into a structure to make the information accessible to the large language model, and will be available to the engine the next morning. The OAuth 2.0 standard is used to link the service to the individual users, providing secure identification for sensitive data. The original data also never leaves the server, so data integrity is ensured and access is strictly controlled.
TARS MCP Servers are custom built to the clients specifications, needs and data. In effect the server gives even junior or new members of your team the ability to quickly access and use institutional knowledge in the way similar to a more senior and long term employee. Through the MCP pipeline the server will ingest the data and automatically categorise and enumerate it for Large Language Model use according to the client specifications. The model will then be able to query and use the data for reasoning. Users can drop new information into the pipeline on a daily basis, the system updates automatically. There are various control and feedback processes in place to adjust and improve the data quality and service continuously.
