Relay
Relay · Productivity
Relay is a browser-first cross-AI context manager for developers. It's an AI memory tool that captures decisions, tasks, and constraints from supported AI chats, turns them into a living project brief, and keeps that brief synced between browser sessions and MCP integration points in coding agents like Claude Code, Cursor, Windsurf, and Codex. Instead of re-explaining your project every time you open a fresh chat, you point the tool at your stored context and keep working. That's the whole pitch.
About Relay
What Is Relay
Relay is a memory sidecar for the AI tools you already use. Its job is to stop the same project context from getting lost every time you start a new chat or switch to a different assistant. Simple idea. Hard to get right.
It watches supported AI chats, pulls out the things that matter (objectives, decisions, constraints, tasks, and risks), and stores them as structured entries you can review. Those entries carry a status such as active, tentative, disputed, or stale, so you can tell current truth from something that's since been overturned. Status labels matter here. Context rots fast.
The big caveat is scope. Relay works through a Chrome extension and an MCP connection, so it leans on your browser and your IDE setup. If you only use one AI tool on one device and rarely start fresh conversations, the payoff is small. It pays off most for people juggling several assistants on the same codebase.
Getting Started
- Sign up at onrelay.app and create your first project.
- Install the Chrome extension if you want Relay to capture context from your browser chats.
- Run
npx @onrelay/wizardto connect the MCP server and configure your IDE. - Open a work session, do your normal back-and-forth with an assistant, then checkpoint what matters.
- Pull the generated brief into a fresh chat or coding agent to pick up where you left off.
Product Information
A quick look at Relay's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Solo developers running several AI assistants on one project
- Engineers who live in an IDE agent
- Small teams standardizing on documented project history
Tasks
- Onboarding an AI into an existing codebase
- Resuming unfinished work
- Tracking decisions over time
Scenarios
- Switching between a browser chat and an IDE agent mid-task, when the second tool would otherwise start from nothing.
- Long-running projects where the original rationale behind an architecture choice has been forgotten.
- Cross-checking conflicting instructions from two different AI surfaces before you commit code.
Key features
Cross-AI context capture
Relay keeps one living project brief that pools AI chat context from multiple assistants. Nothing stays trapped inside a single tool. You add notes and decisions once, and the brief stays available in a browser chat or an IDE. Who wants to retype the same background twice?
MCP integration for coding agents
The MCP server exposes tools like get_brief, recall, save, and list_projects to clients such as Claude Code, Cursor, Windsurf, and Codex. Think of it as coding agent memory that travels with you. Your agent can pull the current brief or save a new decision without you leaving the editor. No copy-paste required.
Observe, reflect, promote cycle
Raw session observations don't become fact on their own. Relay runs them through an observe, reflect, and promote loop, so only reviewed material gets promoted into durable project context. It's a deliberate gate, and it cuts down on half-formed ideas polluting your brief. Nothing worse than a ghost decision in your notes.
Structured context entries
Context is stored as typed entries (objectives, decisions, constraints, tasks, progress, architecture facts, risks, and assumptions), each with a status. You get current-state answers and historical ones, which matters when someone asks why a decision was made. That history is the point. Otherwise you're guessing at your own past reasoning.
Drift reconciliation
When two surfaces make conflicting claims, Relay flags the disagreement for review instead of silently picking a winner. That matters in practice. An IDE agent and a browser chat can drift apart on what the project is supposed to do, and you'd rather know before you commit code.
Browser extension capture
The Chrome extension handles capture from supported AI chats in the browser. Text you'd normally lose to a closed tab gets pulled into the project instead. One less thing to remember.
REST API access
Beyond MCP, Relay exposes a REST API for projects, memory, context briefs, and work sessions. You authenticate with a Bearer token created in settings. That opens the door for scripts and internal tools to read or write context too.
Handoff packets
Relay composes context packets for different targets, such as a fresh chat bootstrap, a quick continuity note, or a full agent bootstrap. You pick the packet that matches how much the next tool needs to know. Small choice, big difference.
Pros and cons
Pros
- Keeps one project brief synced across browser chats and IDE agents, so you stop re-explaining the same background.
- The observe-reflect-promote gate keeps unreviewed guesses out of your durable context.
- Status labels and validity windows let you separate current truth from superseded decisions.
- MCP tools plug into Claude Code, Cursor, Windsurf, and Codex with a single `npx` wizard.
- A free tier with 2 active projects makes it cheap to test on real work before paying.
Cons
- Capture depends on the Chrome extension and an MCP setup, so it isn't a set-and-forget tool for every AI app you use.
- Free-tier limits are tight: 15 reads and 5 writes per day will run out fast on an active project.
- The API rate limits and paid-tier pricing aren't published as fixed numbers in the current docs, so budgeting takes a support email.
Frequently asked questions
It captures decisions, tasks, and constraints from your AI chats and turns them into a shared project brief. You then inject that brief into fresh chats or coding agents so they start with your project's real context instead of a blank slate.
Related content
Explore related tools, skills, and articles for Relay.
Relay Alternatives
Udon
Udon · Coding · ProductivityUdon turns an Apple Silicon Mac into a home server you run from a browser, with container controls, file shares, and an optional AI sysadmin.
offstage
offstage cloud · Coding · Productivityoffstage gives AI coding agents and CI jobs a real Mac: a fresh macOS account on Apple silicon, driven over an API and billed by the second.
BotBus
Chengdu Tuyang Tusen Technology Co., Ltd. · Coding · ProductivityBotBus lets you control AI coding agents from your phone and watch: approve commands, start tasks and read results while the code stays on your desktop.