Build instant multiplayer web apps, no server required
π TRY THE DEMO π
Trystero manages a clandestine courier network that lets your application's users talk directly with one another, encrypted and without a server middleman.
The net is full of open, decentralized communication channels: torrent trackers, IoT device brokers, boutique file protocols, and niche social networks.
Trystero piggybacks on these networks to automatically establish secure, private, p2p connections between your app's users with no effort on your part.
Peers can connect via π BitTorrent, π¦ Nostr, π‘ MQTT, β‘οΈ Supabase, π₯ Firebase, or πͺ IPFS β all using the same API.
Besides making peer matching automatic, Trystero offers some nice abstractions on top of WebRTC:
- ππ£ Rooms / broadcasting
- π’π© Automatic serialization / deserialization of data
- π₯π· Attach metadata to binary data and media streams
- βοΈβ³ Automatic chunking and throttling of large data
- β±π€ Progress events and promises for data transfers
- ππ Session data encryption
- βοΈπͺ React hooks
You can see what people are building with Trystero here.
- How it works
- Get started
- Listen for events
- Broadcast events
- Audio and video
- Advanced
- API
- Strategy comparison
π If you just want to try out Trystero, you can skip this explainer and jump into using it.
To establish a direct peer-to-peer connection with WebRTC, a signalling channel is needed to exchange peer information (SDP). Typically this involves running your own matchmaking server but Trystero abstracts this away for you and offers multiple "serverless" strategies for connecting peers (currently BitTorrent, Nostr, MQTT, Supabase, Firebase, and IPFS).
The important point to remember is this:
π
Beyond peer discovery, your app's data never touches the strategy medium and is sent directly peer-to-peer and end-to-end encrypted between users.
π
You can compare strategies here.
You can install with npm (npm i trystero
) and import like so:
import {joinRoom} from 'trystero'
Or maybe you prefer a simple script tag? Download a pre-built JS file from the latest release and import it locally:
<script type="module">
import {joinRoom} from './trystero-torrent.min.js'
</script>
By default, the Nostr strategy is used. To use a different one just deep import like so (your bundler should handle including only relevant code):
import {joinRoom} from 'trystero/mqtt' // (trystero-mqtt.min.js with a local file)
// or
import {joinRoom} from 'trystero/torrent' // (trystero-torrent.min.js)
// or
import {joinRoom} from 'trystero/supabase' // (trystero-supabase.min.js)
// or
import {joinRoom} from 'trystero/firebase' // (trystero-firebase.min.js)
// or
import {joinRoom} from 'trystero/ipfs' // (trystero-ipfs.min.js)
Next, join the user to a room with an ID:
const config = {appId: 'san_narciso_3d'}
const room = joinRoom(config, 'yoyodyne')
The first argument is a configuration object that requires an appId
. This
should be a completely unique identifier for your appΒΉ. The second argument
is the room ID.
Why rooms? Browsers can only handle a limited amount of WebRTC connections at a time so it's recommended to design your app such that users are divided into groups (or rooms, or namespaces, or channels... whatever you'd like to call them).
ΒΉ When using Firebase, appId
should be your databaseURL
and when using
Supabase, it should be your project URL.
Listen for peers joining the room:
room.onPeerJoin(peerId => console.log(`${peerId} joined`))
Listen for peers leaving the room:
room.onPeerLeave(peerId => console.log(`${peerId} left`))
Listen for peers sending their audio/video streams:
room.onPeerStream(
(stream, peerId) => (peerElements[peerId].video.srcObject = stream)
)
To unsubscribe from events, leave the room:
room.leave()
You can access the local user's peer ID by importing selfId
like so:
import {selfId} from 'trystero'
console.log(`my peer ID is ${selfId}`)
Send peers your video stream:
room.addStream(
await navigator.mediaDevices.getUserMedia({audio: true, video: true})
)
Send and subscribe to custom P2P actions:
const [sendDrink, getDrink] = room.makeAction('drink')
// buy drink for a friend
sendDrink({drink: 'negroni', withIce: true}, friendId)
// buy round for the house (second argument omitted)
sendDrink({drink: 'mezcal', withIce: false})
// listen for drinks sent to you
getDrink((data, peerId) =>
console.log(
`got a ${data.drink} with${data.withIce ? '' : 'out'} ice from ${peerId}`
)
)
You can also use actions to send binary data, like images:
const [sendPic, getPic] = room.makeAction('pic')
// blobs are automatically handled, as are any form of TypedArray
canvas.toBlob(blob => sendPic(blob))
// binary data is received as raw ArrayBuffers so your handling code should
// interpret it in a way that makes sense
getPic(
(data, peerId) => (imgs[peerId].src = URL.createObjectURL(new Blob([data])))
)
Let's say we want users to be able to name themselves:
const idsToNames = {}
const [sendName, getName] = room.makeAction('name')
// tell other peers currently in the room our name
sendName('Oedipa')
// tell newcomers
room.onPeerJoin(peerId => sendName('Oedipa', peerId))
// listen for peers naming themselves
getName((name, peerId) => (idsToNames[peerId] = name))
room.onPeerLeave(peerId =>
console.log(`${idsToNames[peerId] || 'a weird stranger'} left`)
)
Actions are smart and handle serialization and chunking for you behind the scenes. This means you can send very large files and whatever data you send will be received on the other side as the same type (a number as a number, a string as a string, an object as an object, binary as binary, etc.).
Here's a simple example of how you could create an audio chatroom:
// this object can store audio instances for later
const peerAudios = {}
// get a local audio stream from the microphone
const selfStream = await navigator.mediaDevices.getUserMedia({
audio: true,
video: false
})
// send stream to peers currently in the room
room.addStream(selfStream)
// send stream to peers who join later
room.onPeerJoin(peerId => room.addStream(selfStream, peerId))
// handle streams from other peers
room.onPeerStream((stream, peerId) => {
// create an audio instance and set the incoming stream
const audio = new Audio()
audio.srcObject = stream
audio.autoplay = true
// add the audio to peerAudio object if you want to address it for something
// later (volume, etc.)
peerAudios[peerId] = audio
})
Doing the same with video is similar, just be sure to add incoming streams to video elements in the DOM:
const peerVideos = {}
const videoContainer = document.getElementById('videos')
room.onPeerStream((stream, peerId) => {
let video = peerVideos[peerId]
// if this peer hasn't sent a stream before, create a video element
if (!video) {
video = document.createElement('video')
video.autoplay = true
// add video element to the DOM
videoContainer.appendChild(video)
}
video.srcObject = stream
peerVideos[peerId] = video
})
Let's say your app supports sending various types of files and you want to annotate the raw bytes being sent with metadata about how they should be interpreted. Instead of manually adding metadata bytes to the buffer you can simply pass a metadata argument in the sender action for your binary payload:
const [sendFile, getFile] = makeAction('file')
getFile((data, peerId, metadata) =>
console.log(
`got a file (${metadata.name}) from ${peerId} with type ${metadata.type}`,
data
)
)
// to send metadata, pass a third argument
// to broadcast to the whole room, set the second peer ID argument to null
sendFile(buffer, null, {name: 'The CourierΚΌs Tragedy', type: 'application/pdf'})
Action sender functions return a promise that resolves when they're done sending. You can optionally use this to indicate to the user when a large transfer is done.
await sendFile(amplePayload)
console.log('done sending to all peers')
Action sender functions also take an optional callback function that will be continuously called as the transmission progresses. This can be used for showing a progress bar to the sender for large tranfers. The callback is called with a percentage value between 0 and 1 and the receiving peer's ID:
sendFile(
payload,
// notice the peer target argument for any action sender can be a single peer
// ID, an array of IDs, or null (meaning send to all peers in the room)
[peerIdA, peerIdB, peerIdC],
// metadata, which can also be null if you're only interested in the
// progress handler
{filename: 'paranoids.flac'},
// assuming each peer has a loading bar added to the DOM, its value is
// updated here
(percent, peerId) => (loadingBars[peerId].value = percent)
)
Similarly you can listen for progress events as a receiver like this:
const [sendFile, getFile, onFileProgress] = room.makeAction('file')
onFileProgress((percent, peerId, metadata) =>
console.log(
`${percent * 100}% done receiving ${metadata.filename} from ${peerId}`
)
)
Notice that any metadata is sent with progress events so you can show the receiving user that there is a transfer in progress with perhaps the name of the incoming file.
Since a peer can send multiple transmissions in parallel, you can also use metadata to differentiate between them, e.g. by sending a unique ID.
Once peers are connected to each other all of their communications are
end-to-end encrypted. During the initial connection / discovery process, peers'
SDPs are sent via
the chosen peering strategy medium. By default the SDP is encrypted using a key
derived from your app ID and room ID to prevent plaintext session data from
appearing in logs. This is fine for most use cases, however a relay strategy
operator can reverse engineer the key using the room and app IDs. A more secure
option is to pass a password
parameter in the app configuration object which
will be used to derive the encryption key:
joinRoom({appId: 'kinneret', password: 'MuchoMaa$'}, 'w_a_s_t_e__v_i_p')
This is a shared secret that must be known ahead of time and the password must match for all peers in the room for them to be able to connect. An example use case might be a private chat room where users learn the password via external means.
Trystero functions are idempotent so they already work out of the box as React hooks.
Here's a simple example component where each peer syncs their favorite color to everyone else:
import {joinRoom} from 'trystero'
import {useState} from 'react'
const trysteroConfig = {appId: 'thurn-und-taxis'}
export default function App({roomId}) {
const room = joinRoom(trysteroConfig, roomId)
const [sendColor, getColor] = room.makeAction('color')
const [myColor, setMyColor] = useState('#c0ffee')
const [peerColors, setPeerColors] = useState({})
// whenever new peers join the room, send my color to them:
room.onPeerJoin(peer => sendColor(myColor, peer))
// listen for peers sending their colors and update the state accordingly:
getColor((color, peer) =>
setPeerColors(peerColors => ({...peerColors, [peer]: color}))
)
const updateColor = e => {
const {value} = e.target
// when updating my own color, broadcast it to all peers:
sendColor(value)
setMyColor(value)
}
return (
<>
<h1>Trystero + React</h1>
<h2>My color:</h2>
<input type="color" value={myColor} onChange={updateColor} />
<h2>Peer colors:</h2>
<ul>
{Object.entries(peerColors).map(([peerId, color]) => (
<li key={peerId} style={{backgroundColor: color}}>
{peerId}: {color}
</li>
))}
</ul>
</>
)
}
Astute readers may notice the above example is simple and doesn't consider if we
want to change the component's room ID or unmount it. For those scenarios you
can use this simple useRoom()
hook that unsubscribes from room events
accordingly:
import {joinRoom} from 'trystero'
import {useEffect, useRef} from 'react'
export const useRoom = (roomConfig, roomId) => {
const roomRef = useRef(joinRoom(roomConfig, roomId))
useEffect(() => {
roomRef.current = joinRoom(roomConfig, roomId)
return () => roomRef.current.leave()
}, [roomConfig, roomId])
return roomRef.current
}
To use the Supabase strategy:
- Create a Supabase project or use an existing one
- On the dashboard, go to Project Settings -> API
- Copy the Project URL and set that as the
appId
in the Trystero config, copy theanon public
API key and set it assupabaseKey
in the Trystero config
If you want to use the Firebase strategy and don't have an existing project:
- Create a Firebase project
- Create a new Realtime Database
- Copy the
databaseURL
and use it as theappId
in your Trystero config
[*Optional*] Configure the database with security rules to limit activity:
{
"rules": {
".read": false,
".write": false,
"__trystero__": {
".read": false,
".write": false,
"$room_id": {
".read": true,
".write": true
}
}
}
}
These rules ensure room peer presence is only readable if the room namespace is known ahead of time.
Adds local user to room whereby other peers in the same namespace will open
communication channels and send events. Calling joinRoom()
multiple times with
the same namespace will return the same room instance.
-
config
- Configuration object containing the following keys:-
appId
- (required) A unique string identifying your app. When using Supabase, this should be set to your project URL (see Supabase setup instructions). If using Firebase, this should be thedatabaseURL
from your Firebase config (also seefirebaseApp
below for an alternative way of configuring the Firebase strategy). -
password
- (optional) A string to encrypt session descriptions via AES-GCM as they are passed through the peering medium. If not set, session descriptions will be encrypted with a key derived from the app ID and room name. A custom password must match between any peers in the room for them to connect. See encryption for more details. -
rtcConfig
- (optional) Specifies a customRTCConfiguration
for all peer connections. -
relayUrls
- (optional, π BitTorrent, π¦ Nostr, π‘ MQTT only) Custom list of URLs for the strategy to use to bootstrap P2P connections. These would be BitTorrent trackers, Nostr relays, and MQTT brokers, respectively. They must support secure WebSocket connections. -
relayRedundancy
- (optional, π BitTorrent, π¦ Nostr, π‘ MQTT only) Integer specifying how many torrent trackers to connect to simultaneously in case some fail. Passing arelayUrls
option will cause this option to be ignored as the entire list will be used. -
supabaseKey
- (required, β‘οΈ Supabase only) Your Supabase project'sanon public
API key. -
firebaseApp
- (optional, π₯ Firebase only) You can pass an already initialized Firebase app instance instead of anappId
. Normally Trystero will initialize a Firebase app based on theappId
but this will fail if youΚΌve already initialized it for use elsewhere. -
rootPath
- (optional, π₯ Firebase only) String specifying path where Trystero writes its matchmaking data in your database ('__trystero__'
by default). Changing this is useful if you want to run multiple apps using the same database and don't want to worry about namespace collisions. -
libp2pConfig
- (optional, πͺ IPFS only)Libp2pOptions
where you can specify a list of static peers for bootstrapping.
-
-
roomId
- A string to namespace peers and events within a room. -
onError(details)
- (optional) A callback function that will be called if the room cannot be joined due to an incorrect password.details
is an object containingappId
,roomId
,peerId
, anderror
decribing the error.
Returns an object with the following methods:
-
Remove local user from room and unsubscribe from room events.
-
Returns a map of
RTCPeerConnection
s for the peers present in room (not including the local user). The keys of this object are the respective peers' IDs. -
Broadcasts media stream to other peers.
-
stream
- AMediaStream
with audio and/or video to send to peers in the room. -
targetPeers
- (optional) If specified, the stream is sent only to the target peer ID (string) or list of peer IDs (array). -
metadata
- (optional) Additional metadata (any serializable type) to be sent with the stream. This is useful when sending multiple streams so recipients know which is which (e.g. a webcam versus a screen capture). If you want to broadcast a stream to all peers in the room with a metadata argument, passnull
as the second argument.
-
-
Stops sending previously sent media stream to other peers.
-
stream
- A previously sentMediaStream
to stop sending. -
targetPeers
- (optional) If specified, the stream is removed only from the target peer ID (string) or list of peer IDs (array).
-
-
Adds a new media track to a stream.
-
track
- AMediaStreamTrack
to add to an existing stream. -
stream
- The targetMediaStream
to attach the new track to. -
targetPeers
- (optional) If specified, the track is sent only to the target peer ID (string) or list of peer IDs (array). -
metadata
- (optional) Additional metadata (any serializable type) to be sent with the track. Seemetadata
notes foraddStream()
above for more details.
-
-
Removes a media track from a stream.
-
track
- TheMediaStreamTrack
to remove. -
stream
- TheMediaStream
the track is attached to. -
targetPeers
- (optional) If specified, the track is removed only from the target peer ID (string) or list of peer IDs (array).
-
-
Replaces a media track with a new one.
-
oldTrack
- TheMediaStreamTrack
to remove. -
newTrack
- AMediaStreamTrack
to attach. -
stream
- TheMediaStream
theoldTrack
is attached to. -
targetPeers
- (optional) If specified, the track is replaced only for the target peer ID (string) or list of peer IDs (array).
-
-
Registers a callback function that will be called when a peer joins the room. If called more than once, only the latest callback registered is ever called.
callback(peerId)
- Function to run whenever a peer joins, called with the peer's ID.
Example:
onPeerJoin(peerId => console.log(`${peerId} joined`))
-
Registers a callback function that will be called when a peer leaves the room. If called more than once, only the latest callback registered is ever called.
callback(peerId)
- Function to run whenever a peer leaves, called with the peer's ID.
Example:
onPeerLeave(peerId => console.log(`${peerId} left`))
-
Registers a callback function that will be called when a peer sends a media stream. If called more than once, only the latest callback registered is ever called.
callback(stream, peerId, metadata)
- Function to run whenever a peer sends a media stream, called with the the peer's stream, ID, and optional metadata (seeaddStream()
above for details).
Example:
onPeerStream((stream, peerId) => console.log(`got stream from ${peerId}`, stream) )
-
Registers a callback function that will be called when a peer sends a media track. If called more than once, only the latest callback registered is ever called.
callback(track, stream, peerId, metadata)
- Function to run whenever a peer sends a media track, called with the the peer's track, attached stream, ID, and optional metadata (seeaddTrack()
above for details).
Example:
onPeerTrack((track, stream, peerId) => console.log(`got track from ${peerId}`, track) )
-
Listen for and send custom data actions.
actionId
- A string to register this action consistently among all peers.
Returns an array of three functions:
-
-
Sends data to peers and returns a promise that resolves when all target peers are finished receiving data.
-
(data, [targetPeers], [metadata], [onProgress])
-
data
- Any value to send (primitive, object, binary). Serialization and chunking is handled automatically. Binary data (e.g.Blob
,TypedArray
) is received by other peer as an agnosticArrayBuffer
. -
targetPeers
- (optional) Either a peer ID (string), an array of peer IDs, ornull
(indicating to send to all peers in the room). -
metadata
- (optional) If the data is binary, you can send an optional metadata object describing it (see Binary metadata). -
onProgress
- (optional) A callback function that will be called as every chunk for every peer is transmitted. The function will be called with a value between 0 and 1 and a peer ID. See Progress updates for an example.
-
-
-
-
Registers a callback function that runs when data for this action is received from other peers.
-
(data, peerId, metadata)
-
data
- The value transmitted by the sending peer. Deserialization is handled automatically, i.e. a number will be received as a number, an object as an object, etc. -
peerId
- The ID string of the sending peer. -
metadata
- (optional) Optional metadata object supplied by the sender ifdata
is binary, e.g. a filename.
-
-
-
-
Registers a callback function that runs when partial data is received from peers. You can use this for tracking large binary transfers. See Progress updates for an example.
-
(percent, peerId, metadata)
-
percent
- A number between 0 and 1 indicating the percentage complete of the transfer. -
peerId
- The ID string of the sending peer. -
metadata
- (optional) Optional metadata object supplied by the sender.
-
-
Example:
const [sendCursor, getCursor] = room.makeAction('cursormove') window.addEventListener('mousemove', e => sendCursor([e.clientX, e.clientY])) getCursor(([x, y], peerId) => { const peerCursor = cursorMap[peerId] peerCursor.style.left = x + 'px' peerCursor.style.top = y + 'px' })
-
Takes a peer ID and returns a promise that resolves to the milliseconds the round-trip to that peer took. Use this for measuring latency.
peerId
- Peer ID string of the target peer.
Example:
// log round-trip time every 2 seconds room.onPeerJoin(peerId => setInterval( async () => console.log(`took ${await room.ping(peerId)}ms`), 2000 ) )
A unique ID string other peers will know the local user as globally across rooms.
(π BitTorrent, π¦ Nostr, π‘ MQTT only) Returns an object of relay URL keys mapped to their WebSocket connections. This can be useful for determining the state of the user's connection to the relays and handling any connection failures.
Example:
console.log(trystero.getRelaySockets())
// => Object {
// "wss://tracker.webtorrent.dev": WebSocket,
// "wss://tracker.openwebtorrent.com": WebSocket
// }
(π₯ Firebase only) Returns a promise that resolves to a list of user IDs present in the given namespace. This is useful for checking how many users are in a room without joining it.
config
- A configuration objectroomId
- A namespace string that you'd pass tojoinRoom()
.
Example:
console.log((await trystero.getOccupants(config, 'the_scope')).length)
// => 3
one-time setupΒΉ | bundle sizeΒ² | time to connectΒ³ | |
---|---|---|---|
π¦ Nostr | none π | 54K | β±οΈβ±οΈ |
π‘ MQTT | none π | 332K | β±οΈβ±οΈ |
π BitTorrent | none π | 25K π | β±οΈβ±οΈ |
β‘οΈ Supabase | ~5 mins | 150K | β±οΈ π |
π₯ Firebase | ~5 mins | 177K | β±οΈ π |
πͺ IPFS | none π | 945K | β±οΈβ±οΈ |
ΒΉ All strategies except Firebase require zero setup. Firebase is a managed strategy which requires setting up an account.
Β² Calculated via Rollup bundling + Terser compression.
Β³ Relative speed of peers connecting to each other when joining a room. Firebase is near-instantaneous while the other strategies are a bit slower to exchange peering info.
TrysteroΚΌs unique advantage is that it requires zero backend setup and uses decentralized infrastructure in most cases. This allows for frictionless experimentation and no single point of failure. One potential drawback is that itΚΌs difficult to guarantee that the public infrastructure it uses will always be highly available, even with the redundancy techniques Trystero uses. While the other strategies are decentralized, the Supabase and Firebase strategies are a more managed approach with greater control and an SLA, which might be more appropriate for βproductionβ apps.
Trystero makes it trivial to switch between strategies β just change a single import line and quickly experiment:
import {joinRoom} from 'trystero/[torrent|nostr|mqtt|supabase|firebase|ipfs]'
Trystero by Dan Motzenbecker