IRCv3 Specifications

IRCv3 specifications build on top of the core IRC protocol. The primary core is described by the “Modern IRC” specification, which supersedes historical IRC protocol specifications (RFC1459, RFC2812 and RFC7194). To fully understand IRCv3, please read the “Modern IRC” specification followed by the IRCv3 specifications.

The IRCv3 specifications are released when they are stable and have been widely tested. In the past the WG released specifications as versioned bundles (IRCv3.1, IRCv3.2), but we no longer do this.

Errata updates may be submitted for our specifications. To do so, simply see our contribution document.

Capability Negotiation 🔗

Capabilities let us implement protocol changes in backwards-compatible ways. They also convey various information on joining a server. Capability negotiation is a vital part of IRCv3, and lets clients request and enable the CAPs they support once they’ve joined a server.

The Capability Negotiation spec conveys the basic listing and requesting of capabilities, and lays the framework which most IRCv3 specs use. It also goes over the 302 extensions, and cap-notify – a feature to let clients know when capabilities are added to or removed from the server (for example, if the SASL authentication layer disconnects, the associated capability may be disabled for a time).

Message Tags 🔗

Message tags extend the core framing used with IRC messages, and allow extended data to be sent with messages.

Message tags are widely used in the IRCv3 specifications. As such, most software implementing IRCv3 extensions will want to to implement the core Message Tags specification.

The Message Tags spec covers the new message format, how tag data is formatted and escaped, and how they are named. In addition, it extends the message length and lets clients send tags directly between each other, allowing new features to be developed and implemented independently from the IRC servers themselves (similar to extensions based on CTCP).

Note: Message tags themselves are used as a foundation for other extensions and do not themselves offer any user-facing features. Specific message tags are defined in the various IRCv3 specifications.

Account Authentication and Registration 🔗

Accounts have become a first-class feature on IRC. They store channel access and ownership information, preferences, settings, nickname ownership, and even more. Our extensions describe simple account creation and authentication.

The work-in-progress Account Registration spec lets clients build nice account registration interfaces, instead of making users manually send messages to service bots.

SASL lets users authenticate in a standardised way across different IRC networks. It gives a way to authenticate while connecting, without having to message service bots after they’ve joined the server. Because SASL allows authentication before they’ve finished connecting, it lets clients join private and restricted channels without having to setup complex join-wait systems.

The v3.1 SASL spec defines the AUTHENTICATE command and sasl cap, which work together to allow clients to authenticate to the network.

The v3.2 SASL spec defines a way to advertise the authentication methods available to clients, allows for clients to re-authenticate after services is lost and reconnects, and defines what to do if the authentication layer is disconnected or reconnected.

IRC SASL authentication primarily uses the same mechanisms as SASL in other protocols. Most commonly:

For further information on SASL mechanism support, see the SASL Mechanisms page.

Account Tracking 🔗

IRCv3 extensions allow clients to much more easily know when other users are logged into accounts. This allows for much greater integration between client bots and the network’s authentication system, as well as better general display and authentication of client identities.

The account-extban spec defines an ISUPPORT token allowing clients to construct an EXTBAN targeting a specific user account, e.g. to prevent them from joining a channel, make it exempt from bans, or permanently allow it to an invite-only channel.

The account-notify spec defines a way for clients to be notified when other clients login to accounts. This spec defines the ACCOUNT message to enable this, use of the a WHOX token, as well as outlining the general restriction of account names not being * (as this is used to indicate logging out of accounts).

The account-tag spec defines a way for clients to receive a message tag on messages specifying the current account that other client is logged into (or that they aren’t logged into one at all). This is especially useful for letting bots make use of the network’s authentication and account mechanisms.

The extended-join spec defines a way to request that extra client information (including that client’s account) is sent when clients join a given channel. This allows better tracking of accounts, particularly when used with account-notify.

Away Notifications 🔗

The away-notify extension provides a way for clients to instantly know when other clients go away or come back. This improves responsiveness and the display of channels for IRC clients that display this information.

The away-notify spec describes how to sign up for these notifications and the AWAY message to enable this.

Batches 🔗

The batch extension provides a way for servers to mark certain messages as related. This can simplify the display of this information in clients as well as allow better post-processing on them.

The batch spec describes the naming of new batch types, the semantics of batches and how clients should process them.

The work-in-progress client initiated batch extension describes client-to-server batches.

Note: Batches themselves are used as a foundation for other extensions and do not themselves offer any user-facing features.

Here are the standalone batch types the IRCv3 WG defines:

Bot Mode 🔗

The Bot mode lets bots mark themselves as such. Other users will see the client as a bot in various ways, and should see a tag on that client’s messages if they have requested the message-tags capability.

The Bot Mode spec describes the mode and how to see bots that have set the mode on themselves.

Channel renaming 🔗

The channel-rename extension adds a new command sent by clients and servers that enables renaming a channel without closing it down and redirecting to a new one.

The work-in-progress channel-rename spec describes how to use the RENAME command to achieve this.

Persistence 🔗

The work-in-progress chathistory spec describes the syntax and semantics of the new CHATHISTORY command, which standardizes a mechanism for clients to request message history from servers or bouncers.

The work-in-progress message-redaction spec adds a new command to remove a message from the history, and indicate to other clients they should hide it.

The work-in-progress read-marker spec adds a new command to synchronize read markers between several clients of the same user.

The work-in-progress pre-away spec allows clients to send AWAY commands during connection registration.

Changing User Properties 🔗

Several extensions allow servers to notify clients of properties of other users that were historically considered nonchanging for the lifetime of a connection.

The chghost spec describes the new CHGHOST message which lets clients more easily see when other clients’ usernames and/or hostnames are changed. This replaces the clunky method of sending a fake QUIT, and then one or more fake JOIN messages instead.

The setname spec describes the new SETNAME message which allows clients to update their realname (gecos) after connecting to the server, and see updates from other users. This is especially useful as some clients use the realname information for avatars and an additional name field.

Client-Only Tags 🔗

Client-only tags are message tags that are sent directly between clients with no server involvement. They’re special in IRCv3 as they only apply to clients, and as such we detail them in their own section here.

Here are the client-only tags the IRCv3 WG defines:

Echo Message 🔗

The echo-message extension lets clients confirm when messages are sent, and see messages that other clients on their connection (say, via an IRC bouncer) have sent. It does this by echoing messages back to clients after they are sent, allowing for these extra features.

The echo-message spec describes which messages are echo’d, and how they are interpreted by clients.

Invite Notify 🔗

The invite-notify extension allows privileged channel users to see when someone is invited to their channel. This can help chanops better run their channels and see better information about what’s going on.

The invite-notify spec describes the new INVITE reply which this extension uses, and how clients interpret these notifications.

Labeled Responses 🔗

The labeled-response extension allows clients to link returned numerics with sent commands. This allows for much richer/accurate implementations of echo-message, and lets clients generally corrolate sent and received messages.

Additionally, this should also assist bouncers with correctly directing responses to the right connected client.

The labeled-response spec covers the label tag, and how clients should send and will receive labels.

Listing Users 🔗

IRCv3 specifications extend traditional queries (NAMES, WHOIS, WHO) to carry more information about other users in a given channel.

The multi-prefix spec details changes to these queries, which allow clients to see all the statuses (i.e. voice, chanop) that other clients have in a channel rather than just the highest. This improves data tracking for clients and bots, and allows clients to display the privilege level of other clients more correctly.

The userhost-in-names spec describes how the NAMES message changes with this capability active, and how clients should interpret the changes. This allows clients to more easily see the user/hostnames of other clients when joining channels. This allows clients to better track info and automate client features more easily.

The WHOX spec describes how the WHO message and its replies changes with this capability active to allow clients to request more data, and how clients should interpret these changes.

The work-in-progress no-implicit-names spec allows clients to disable the implicit NAMES responses sent after JOIN in case they don’t always need that information for all channels. Clients can still query that information as needed via the NAMES or WHO command.

Message IDs 🔗

The message-ids extension allows servers to provide a network-unique identifier on messages (including PRIVMSG/NOTICE). This allows clients to build new features that refer to specific messages, with the knowledge that these identifiers will be unique.

The message-ids spec covers the msgid tag, how servers should generate message IDs and how clients should treat them.

Note: Message IDs themselves are used as a foundation for other extensions and do not themselves offer any user-facing features. Specific IRCv3 extensions will note their use of (and dependency on) message IDs.

Monitor 🔗

The MONITOR command acts as a standardized way for clients to be alerted when other clients enter or exit the network. This is in opposition to ISON, which does this through polling, and WATCH, which differs between vendor implementations.

The Monitor spec details this command, the relevant RPL_ISUPPORT token and the commands used with it.

The Extended Monitor spec builds upon the Monitor spec, and extends it to various events.

Multiline messages 🔗

The multiline extension adds a new batch type and tag sent by clients and servers to send messages that can exceed the usual byte length limit and that can contain line breaks.

The work-in-progress multiline spec describes how to use the draft/multiline batch type and draft/multiline-concat tag to achieve this.

Server Time 🔗

The server-time extension allows clients to see the exact time that messages were sent and received. This allows bouncers to replay information with more accurate time tracking.

The server-time spec describes the time tag, how to specify timestamps and how clients should parse incoming timestamps.

Server Name Indication (SNI) 🔗

SNI makes it easier for servers to send the correct TLS certificate to connecting clients.

The work-in-progress SNI spec provides guidelines for clients and servers, allowing them to better detect the TLS certificate to send based on the server’s hostname.

Standard Replies 🔗

Standard Replies establishes a clean way to send notes, warnings, and errors to users. This helps server/bouncer developers avoid creating new (possibly conflicting) numerics, and ensures clients will be able to display these messages to users nicely.

The standard replies spec describes the format of these messages, how to use them, and guidance on creating useful descriptions for users.

Strict Transport Security (STS) 🔗

STS allows clients to be automatically upgraded to use TLS encryption and prevents downgrade attacks. STS is intended as a replacement for the STARTTLS command with better security qualities.

The sts spec describes the sts capability, how it operates, and various implementation details for both clients and servers.

UTF8ONLY 🔗

The UTF8ONLY ISUPPORT token lets servers indicate that they only support UTF-8 traffic, allowing clients to set their incoming/outgoing encodings automatically.

The UTF8ONLY spec details the RPL_ISUPPORT token, associated messages, and functionality.

WebIRC 🔗

The WEBIRC command is widely used to provide the real IP address of users to the server when connecting through a gateway. This is common for current web-based IRC clients.

The WEBIRC spec describes how this command works, how to use it, and some best practices to keep in mind while implementing this feature.

WebSocket 🔗

The work-in-progress WebSocket spec describes conventions for transporting IRC lines over the WebSocket protocol. This is necessary for browser-based clients, which cannot make conventional TCP connections to IRC servers.

Deprecated Extensions 🔗

These extensions have been explicitly deprecated. We no longer recommend implementing them. Generally, these extensions have either been superseded, or other major implementation issues have been discovered with them.

v3.2 Metadata 🔗

The v3.2 METADATA command was found to have issues related to rate-limiting and excessive notifications, which made it impossible for servers in widespread use to implement. A new Metadata specification is being written to address these issues and overhaul the notification system, so we do not recommend implementing this spec.

STARTTLS 🔗

STARTTLS allows clients to upgrade their plaintext connections to use TLS encryption. In alignment with RFC8314, it is recommended that IRC networks use listeners designed for implicit TLS (such as those that operate on port 6697) and clients instead implement STS support.

The tls spec is still available for reference. It describes how the STARTTLS command works, as well as how connection registration is changed by the introduction of this capability.