Memory

Connector Overview

The Memory gives AI agents high-certainty retrieval from large, complex, semi-structured data, with source traceability on every output. As a dedicated connector within the Agentic Framework, it enables agents to recall any relevant information introduced during a session and return it with a verifiable reference to the original source material.

It is built for settings where generic memory systems are not reliable enough. Every session is scoped to the active source of truth, such as a patient record or EHR, and every retrieved answer is tied back to the exact document, entry, or data point used. Information gathered through tool calls and external systems remains available throughout the session, allowing the agent to build a richer and more complete picture over time.

By embedding Memory into agent workflows, teams can deploy AI systems that retrieve accurately, maintain context across complex interactions, and produce outputs that are auditable, safe, and usable in clinical production.

Capabilities

• High-certainty retrieval from large and messy datasets
• Session-wide recall across documents, structured data, and tool outputs
• Source attribution on every answer
• Single-session, single-record context isolation
• Retrieval quality aligned to needle-in-a-haystack standards

Example Use Cases

• Conversational querying of longitudinal patient records
• Pre-visit clinical summarization with persistent follow-up recall
• Retrieval of referrals, labs, medications, and consult history
• Audit-ready traceability for compliance and regulatory workflows
• Live agent workflows that retain access to mid-session tool outputs