DeepJudge, the institutional intelligence platform for legal AI, introduced the Agent Transfer Protocol, an open protocol designed to enable continuity of context between AI systems.
Increasingly, people are turning to various AI products and experiences to complete their work. While existing protocols, such as the Model Context Protocol (MCP), allow systems to exchange information and perform specific actions, the broader context needed to continue a task—including goals, supporting materials, conversation history, and previous work—is often lost in the process.
The Agent Transfer Protocol addresses this challenge by enabling AI platforms to transfer the user and the context needed to continue work from one system to another. Instead of rebuilding context each time they switch products, users can move between AI experiences while preserving the information and progress that brought them to their current point.
For example, a lawyer investigating a precedent issue in DeepJudge can move on to Harvey with the relevant context already established, allowing them to continue their work without having to reload materials, recreate prompts, or reconstruct the previous analysis.
“No single AI experience will be right for every task,” said Yannic Kilcher, chief technology officer and co-founder of DeepJudge. “People should be able to use the AI tool that best suits their needs and bring their context with them. That’s the experience we created to facilitate with the Agent Transfer Protocol.”
DeepJudge developed the Agent Transfer Protocol and is releasing it as an open protocol. Harvey, the leading legal operating system for legal and professional services, has announced an integration implementing the Agent Transfer Protocol, which will enter beta this month.
Thomson Reuters also supports the Agent Transfer Protocol, and further details on its implementation and timeline will be announced soon. Joel Hron, Chief Technology Officer at Thomson Reuters, said: “The future of professional AI will not be defined by a single model, agent, or application. Professionals will increasingly work with specialized AI experiences, and these systems must collaborate without sacrificing the context, trust, and expertise needed for high-responsibility tasks. Open standards like the Agent Transfer Protocol can facilitate this interoperability, and we at Thomson Reuters are excited to support this initiative and help shape the future.”
“The most efficient legal workflows rely on more than one specialized system. The Agent Transfer Protocol allows a user to move seamlessly between Harvey and a partner’s product without losing track of their work, maximizing the potential of both products. We are pleased to be among the first to implement it,” said Siva Gurumurthy, Chief Technology Officer at Harvey.
As an institutional intelligence platform for the legal field, DeepJudge offers the Agent Transfer Protocol as a shared resource, rather than an exclusive advantage. The company invites other AI providers, specialized platforms, and law firms developing in-house AI solutions to adopt the protocol and contribute to its ongoing development.
The agent transfer protocol specification and the resources for its implementation are available on GitHub.
