Relay

What is Relay?

Relay is a context management tool for people using multiple AI chat and coding environments. It captures decisions, tasks, and constraints from your AI chats and keeps a living “project brief” synced so you can resume work without manually copying transcripts.

The product is designed to work across browser-based chat experiences (including ChatGPT, Claude, Gemini, and others) and IDE coding agents through MCP (Model Context Protocol), where agents can read and write structured project memory.

Key Features

• Auto-capture project context from AI chats: Relay extracts what matters—specifically decisions, tasks, and constraints—quietly as you chat, reducing manual saving.
• One-click restoration via “project briefs”: Start a fresh conversation with the full project brief injected automatically, reflecting updates as you continue working.
• Cross-surface sync between tools: Changes made in one environment (e.g., a choice in a browser chat) can surface in another (e.g., your IDE workflow).
• MCP integration for IDE agents: Relay implements an MCP bridge so coding agents can read the latest brief and write back decisions and progress.
• Works with many AI chat tools: The site lists compatibility with ChatGPT, Claude, Gemini, Grok, Perplexity, DeepSeek, and others, plus support for MCP-connected browser/IDE workflows.

How to Use Relay

1. Get started with Relay: Install the browser extension and/or use the provided setup workflow to connect Relay to your sessions and supported tools.
2. Begin working in your AI chats: As you chat, Relay auto-captures relevant context into your project brief.
3. Open a fresh chat when you need to continue: Use the brief restoration flow so your next conversation starts with the updated project context.
4. For coding agents, connect via MCP: Ensure your IDE agent is MCP-compatible; the agent can then read the latest brief and write back updates.

Use Cases

• Resuming a multi-step build across chat sessions: You make decisions and define constraints in one ChatGPT/Claude/Gemini session, then start a new chat and restore the updated brief to continue.
• Keeping browser chats aligned with an IDE agent: A decision made in a browser chat appears in your IDE agent’s next step (and constraints added in the IDE remain consistent in later browser work).
• Maintaining a “single source of truth” for ongoing projects: Instead of managing separate notes, you rely on Relay’s living project brief that updates as new tasks and constraints emerge.
• Structured memory for agent-driven workflows: An MCP-connected coding agent reads the current brief (e.g., “My App Project”) and writes back decisions and progress during iterative development.
• Multi-tool context switching: You alternate among different AI tools (browser chat and coding agents) while keeping project-level context synchronized rather than repeating the same background.

FAQ

• What does Relay capture from my AI conversations? Relay captures decisions, tasks, and constraints from the conversations it supports.

• How does Relay help when I start a new chat? It supports one-click context restoration by injecting the full project brief into a fresh conversation.

• What is MCP in this context? The site describes MCP (Model Context Protocol) as an open standard for AI agents to read and write structured memory, and Relay uses MCP as a bridge between browser chats and IDE agents.

• Which tools does Relay support? The page lists browser chat support for multiple AI tools (including ChatGPT, Claude, Gemini, Grok, Perplexity, and DeepSeek) and shows an MCP integration path for IDE agents.

• Does Relay require manual context saving? The site describes an auto-capture workflow that is meant to reduce manual saving and copying of transcripts.

Alternatives

• Manual note-taking + copy/paste of summaries: Using a document or knowledge base and pasting key background into each chat is a straightforward alternative, but it typically requires more manual upkeep.
• Standalone project wikis or documentation tools: Keeping a project brief in a wiki/markdown repo and referencing it during chats can provide persistence, but it won’t automatically extract and sync decisions from conversations.
• MCP-compatible agent memory tools (without cross-chat context capture): Other MCP-based systems can support structured memory for agents, but they may not provide Relay’s specific auto-capture and cross-surface brief syncing across browser chat sessions.
• Single-environment “context” features within one AI tool: Some platforms offer their own persistent chat context. This can work within one product, but it won’t address the workflow where context must transfer between multiple chat tools and IDE agents.