Real Time Messaging API

The Real Time Messaging API is a WebSocket-based API that allows you to receive events from Slack in real time and send messages as users. It's sometimes referred to as simply the "RTM API".

It is the basis for all Slack clients. It's also commonly used with the bot user integration to create helper bots for your workspace.

If you prefer events to be pushed to you instead, we recommend using the HTTP-based Events API instead. Most event types supported by the RTM API are also available in the Events API.

New Slack apps may not use any Real Time Messaging API method. Create a classic app and use the V1 Oauth flow to use RTM.

Basics

To begin a RTM session make an authenticated call to the rtm.connect API method. This provides an initial set of workspace metadata and a message server WebSocket URL. Once you have connected to the message server it will provide a stream of events, including both messages and updates to the current state of the workspace. This allows a client to easily maintain a synchronized local copy of all workspace data and messages.

The Websocket URLs provided by rtm.connect are single-use and are only valid for 30 seconds, so make sure to connect quickly. If you connect successfully the first event received will be a hello:

{
    "type": "hello"
}

This will be followed by any events that occurred between the call to rtm.connect and the connection to the message server. If you're reconnecting after a network problem this initial set of events may include a response to the last message sent on a previous connection (with a reply_to) so a client can confirm that message was received.

If there was a problem connecting an error will be returned, including a descriptive error message:

{
    "type": "error",
    "error": {
        "code": 1,
        "msg": "Socket URL has expired"
    }
}

Events

Almost everything that happens in Slack will result in an event being sent to all connected clients. The simplest event is a message sent from a user:

{
    "type": "message",
    "ts": "1358878749.000002",
    "user": "U023BECGF",
    "text": "Hello"
}

Every event has a type property which describes the type of event. Our servers currently send the following event types:

New event types will be added in the future, clients should be able to handle unexpected event types.

Sending messages

You can send a message to Slack by sending JSON over the WebSocket connection.

Every event should have a unique (for that connection) positive integer ID. All replies to that message will include this ID allowing the client to correlate responses with the messages sent; replies may be "out of order" due to the asynchronous nature of the message servers.

Also, as with events sent from the server, each event sent by the client has a string type specifying what the message does — chat messages are of type message.

Channel selection

So to post the text "Hello world" to a channel, you can send this JSON:

{
    "id": 1,
    "type": "message",
    "channel": "C024BE91L",
    "text": "Hello world"
}

You can send a message to a private group or direct message channel in the same way, but using a group ID (C024BE91L) or direct message channel ID (D024BE91L).

To send a message both as and to the authenticating user, use the correct direct message channel ID for that user. The direct message ID can be found as part of the response to rtm.connect, or by consulting conversations.list.

Formatting messages

The RTM API only supports posting simple messages formatted using our default message formatting mode. It does not support attachments or other message formatting modes. To post a more complex message as a user clients can call the chat.postMessage Web API method with as_user set to true.

User mentions over RTM should use the User ID-based <@U123ABC> syntax:

{
    "id": 2,
    "type": "message",
    "channel": "C024BE91L",
    "text": "Hello <@U123ABC>"
}

Handling responses

Once the JSON has been sent to the server visual clients should immediately display the text in the channel, grayed out or otherwise marked to indicate that it is "pending". At some point after that, usually a few milliseconds later, the server will send a confirmation that the message was received:

{
    "ok": true,
    "reply_to": 1,
    "ts": "1355517523.000005",
    "text": "Hello world"
}

Replies to messages sent by clients will always contain two properties: a boolean ok indicating whether they succeeded and an integer reply_to indicating which message they are in response to.

In the case of a reply to a chat message, if successful, the reply will contain the canonical recorded timestamp of the message. All messages within a single channel are guaranteed to have a unique timestamp which is ASCII sortable. Given the precision of the timestamp, clients should treat these timestamps as strings, not floats/doubles. Once a successful reply has been returned, the message in the chat log should no longer be grayed out - it has now been delivered.

Chat message replies also contain the message text, which may vary from the sent message due to URL detection.

Errors

If there is an error processing an event the message server will reply with an error. For example:

{
    "ok": false,
    "reply_to": 1,
    "error": {
        "code": 2,
        "msg": "message text is missing"
    }
}

Typing indicators

Clients can send a typing indicator to indicate that the user is currently writing a message to send to a channel:

{
    "id": 1,
    "type": "typing",
    "channel": "C024BE91L"
}

This can be sent on every key press in the chat input unless one has been sent in the last three seconds. Unless there is an error the server will not send a reply, but it will send a "user_typing" event to all workspace members in the channel.

Presence

User and bot user presence on the RTM API is complicated enough to warrant an entire document. Learn all about presence subscriptions and batch presence events here.

Ping and Pong

Clients should try to quickly detect disconnections, even in idle periods, so that users can easily tell the difference between being disconnected and everyone being quiet. Not all web browsers support the WebSocket ping spec, so the RTM protocol also supports ping/pong messages. When there is no other activity clients should send a ping every few seconds. To send a ping, send the following JSON:

{
    "id": 1234, // ID, see "sending messages" above
    "type": "ping",
    ...
}

You can supply any number of extra "flat" arguments (that is: only scalar values, no arrays or objects). These will be included in the pong message that is sent back. For example, a client could include a local timestamp in the ping message so it can calculate round-trip latency:

{
    "id": 1234,
    "type": "ping",
    "time": 1403299273342
}

This will be included in the reply from the server:

{
    "reply_to": 1234,
    "type": "pong",
    "time": 1403299273342
}

Limits

The message server will disconnect any client that sends a message longer than 16 kilobytes. This includes all parts of the message, including JSON syntax, not just the message text. Clients should limit messages sent to channels to 4000 characters, which will always be under 16k bytes even with a message comprised solely of non-BMP Unicode characters at 4 bytes each. If the message is longer a client should prompt to split the message into multiple messages, create a snippet or create a post.

As with all Slack APIs, the RTM API is subject to rate limits. Clients should not send more than one message per second sustained. If you do you may receive an error message or be disconnected.

What's a WebSocket?

WebSockets are a standard way to open a long-lived bi-directional communication channel with a server over TCP. It's the protocol used when connecting to our RTM API. Many contributions from our community support the particulars of connecting to Slack via a WebSocket.

Connecting with rtm.connect vs. rtm.start

There are currently two ways to reserve websocket connections.

rtm.connect concerns itself only with getting your app connected to the RTM API, and only includes limited information about the connecting user and housing workspace.

rtm.start includes not only an entire kitchen sink but an entire kitchen filled with information about the user, its workspace, its channels, its current state in the universe. rtm.start is naturally more difficult to use with Enterprise Grid and other large workspaces.

We strongly recommend using rtm.connect to reserve your websocket connections and use the Web API in tandem to collect all the state information your app needs.

Events API vs. RTM API

Both APIs are useful for different reasons, despite them both serving the same kind of data. Find out why to choose one or the other (or both!) in the Events API FAQ.

Create a Classic Slack App

Since new Slack apps don't connect to rtm.connect, you'll need to have a classic Slack app to get started. If you haven't created one yet, click here to make one:

Create a classic Slack app