Use cases

Mercure is a thin protocol; it covers a wide range of "I need to push something to a connected client" problems. The pages below are concrete walkthroughs, each one ships a working example you can run.

Modern AI workloads

• LLM token streaming: stream tokens from a server-side OpenAI / Anthropic / local-model call to the browser as they arrive, without a WebSocket gateway in front of your inference server.

• AI agent progress: push state changes from a long-running agent ("searching the web", "running tool", "summarizing") to the UI in real time.

Application real-time use cases

• Live data and dashboards: stock tickers, availability counters, IoT telemetry, observability dashboards.

• Collaborative editing: multiple users edit the same document, changes broadcast as they happen.

• Async jobs and progress: kick off a long-running job, push progress to the requester, deliver the result when ready.

• Notifications: in-app toasts, mention badges, mailbox counters.

Server-rendered apps with Mercure

• Hotwire / Turbo Streams: stream HTML fragments to swap into the page, no JSON layer required.

API integrations with Mercure

• GraphQL subscriptions: back GraphQL subscriptions with Mercure instead of WebSockets.

Mercure in production: case studies

Mercure is used at scale today, a few public examples:

• mail.tm pushes 8M Mercure notifications a day.

• iGraal serves 100k concurrent Mercure users.

• Raven Controls used Mercure to power Cop 21 and Euro 2020.

Don't see your case?

Mercure is the right answer when "the server has fresh data, push it to clients" is the shape of the problem. If you're unsure, ask in GitHub Discussions, most "should I use Mercure for X?" questions have already been answered there.