Relay is a context management system designed for AI interactions, particularly for engineers working across multiple AI tools. It was developed to eliminate the need for users to repeatedly provide the same context to different AI models by maintaining a synchronized, living project brief. Relay ensures that decisions, tasks, and constraints captured from AI chats are consistently available across various platforms, from browser-based AI services to integrated development environment (IDE) agents.

The tool functions by observing conversations with supported AI platforms, automatically extracting key information such as decisions, tasks, and constraints. This extracted data is then used to update a central "project brief." Through its Model Context Protocol (MCP) integration, Relay facilitates a two-way synchronization of this memory between browser-based AI chats like ChatGPT, Claude, and Gemini, and IDE-integrated AI agents such as Claude Code, Cursor, and GitHub Copilot. This ensures that a continuous, up-to-date context is available regardless of the AI interface being used.

Some of the key features are:

• Auto-capture: Automatically saves important information like decisions, tasks, and constraints from AI chat conversations without requiring manual input.
• Project Briefs: Maintains a self-updating project brief that consolidates all relevant context, allowing for one-click restoration of full project history in any new AI chat or session.
• MCP Integration: Enables AI coding agents, including Claude Code and Cursor, to directly read from and write to the shared project memory, aligning their context with browser-based interactions.
• Cross-Surface Sync: Ensures that decisions made in one AI environment, such as a browser chat, are automatically reflected and synchronized with other tools, like an IDE agent, maintaining a unified project state.
• Context Flow: Provides bidirectional context synchronization, allowing decisions made by IDE agents to update the browser brief and vice-versa.
• Truth-scored Governance: Utilizes a system to merge captured data into project memory based on confidence scores, managing potential conflicts and tracking the evolution of project facts.
• Drift Reconciliation: Identifies and flags contradictions or inconsistencies in facts that emerge from simultaneous work across different AI tools or sessions.

Relay's operation centers around a continuous "observe → reflect → promote" cycle. As a user interacts with any supported AI tool, the Relay browser extension or MCP integration quietly monitors the conversation to identify and extract decisions, tasks, and constraints. This extracted information is then sent to Relay's backend, where it is reflected against existing project context. Based on confidence scores, evidence count, and reaffirmation frequency, new facts are promoted to "active" status within the project brief, or conflicts are flagged for review. This process updates the project brief across all linked sessions and tools. When initiating a new AI conversation, the user can easily inject the latest project brief, ensuring the AI has comprehensive context from the outset, thus eliminating the need for repetitive information transfer. The setup typically involves installing a Chrome extension for browser integration and running an npx @onrelay/wizard command for MCP-compatible IDE agents.

Some common use cases include:

• Seamless AI Development: Engineers can ensure consistent project context across diverse AI models and development environments, streamlining AI-assisted coding tasks and maintaining continuity in their workflow.
• Maintaining Consistent Project State: Individuals and teams can rely on Relay to keep project decisions, tasks, and constraints uniformly updated and accessible across all AI tools, significantly reducing the effort of repetitive context explanations.
• Efficient Context Switching: Developers can fluidly transition between browser-based AI chats and IDE-integrated AI agents without losing crucial conversational history or project context, enhancing productivity.
• Managing Complex Software Projects: Relay aids in centralizing and synchronizing architectural facts, identified risks, underlying assumptions, and active tasks that emerge from AI interactions, providing a structured overview for complex projects.
• Onboarding New AI Sessions: Automatically equips new AI chat sessions or coding agents with a comprehensive and up-to-date project brief, eliminating the manual overhead of setting context and allowing immediate productive engagement.