OpenAI has published an overview of V7, a system it describes as giving AI agents "institutional memory." The company says the tool is built on GPT-5.6.
According to OpenAI, V7 addresses a common enterprise problem: company knowledge is spread across many scattered files. V7 turns that fragmented material into context that AI agents can actually use.
With that context in place, OpenAI says agents are able to complete complex work and produce results that are source-linked, meaning outputs can be traced back to the underlying documents.
The announcement frames institutional memory as a foundation for agentic work inside organizations, rather than a standalone retrieval feature.