mbsync - synchronize IMAP4 and Maildir mailboxes


mbsync [options ...] {{channel[:box[{,|\n}...]]|group} ...|-a}


mbsync is a command line application which synchronizes mailboxes; currently Maildir and IMAP4 mailboxes are supported. New messages, message deletions and flag changes can be propagated both ways; the operation set can be selected in a fine-grained manner.
Synchronization is based on unique message identifiers (UIDs), so no identification conflicts can occur (unlike with some other mail synchronizers). OTOH, mbsync is susceptible to UID validity changes (but will recover just fine if the change is unfounded). Synchronization state is kept in one local text file per mailbox pair; these files are protected against concurrent mbsync processes. Mailboxes can be safely modified while mbsync operates (see INHERENT PROBLEMS below for a minor exception). Multiple replicas of each mailbox can be maintained.


-c, --config file

Read configuration from file. By default, the configuration is read from ~/.mbsyncrc.

-a, --all

Select all configured channels. Any channel/group specifications on the command line are ignored.

-l, --list

Don’t synchronize anything, but list all mailboxes in the selected channels and exit.

-C[m][s], --create[-master|-slave]

Override any Create options from the config file. See below.

-R[m][s], --remove[-master|-slave]

Override any Remove options from the config file. See below.

-X[m][s], --expunge[-master|-slave]

Override any Expunge options from the config file. See below.

{-n|-N|-d|-f|-0|-F}, {--new|--renew|--delete|--flags|--noop|--full}
{-L|-H}[n][N][d][f], {--pull|--push}[-new|-renew|-delete|-flags]

Override any Sync options from the config file. See below.

-h, --help

Display a summary of command line options.

-v, --version

Display version information.

-V, --verbose

Enable verbose mode, which displays what is currently happening.


Enable debugging categories:

C, crash - use built-in crash handler
, driver - print driver calls (metadata only)
, driver-all - print driver calls (including messages)
, maildir - print maildir debug info
, main - print main debug info
, net - print network traffic (protocol only)
, net-all - print network traffic (including payloads)
, sync - print synchronization debug info

All categories except crash implictly enable verbose mode. Without category specification, all categories except net-all are enabled.

-q, --quiet

Suppress progress counters (this is implicit if stdout is no TTY, or any debugging categories are enabled) and notices. If specified twice, suppress warning messages as well.


The configuration file is mandatory; mbsync will not run without it. Lines starting with a hash mark (#) are comments and are ignored entirely. Configuration items are keywords followed by one or more arguments; arguments containing spaces must be enclosed in double quotes ("), and literal double quotes and backslashes (\) must be backslash-escaped. All keywords (including those used as arguments) are case-insensitive. Bash-like home directory expansion using the tilde (~) is supported in all options which represent local paths. There are a few global options, the rest applies to particular sections. Sections are started by a section keyword and are terminated by an empty line or end of file. Every section defines an object with an identifier unique within that object class.

There are two basic object classes: Stores and Channels. A Store defines a collection of mailboxes; basically a folder, either local or remote. A Channel connects two Stores, describing the way the two are synchronized.
There are two auxiliary object classes: Accounts and Groups. An Account describes the connection part of remote Stores, so a server connection can be shared between multiple Stores. A Group aggregates multiple Channels to save typing on the command line.

File system locations (in particular, Path and Inbox) use the Store’s internal path separators, which may be slashes, periods, etc., or even combinations thereof.
Mailbox names, OTOH, always use canonical path separators, which are Unix-like forward slashes.

All Stores
These options can be used in all supported Store types.
In this context, the term "remote" describes the second Store within a Channel, and not necessarily a remote server.
The special mailbox INBOX exists in every Store; its physical location in the file system is Store type specific.

The location of the Store in the (server’s) file system. If this is no absolute path, the reference point is Store type specific. This string is prepended to the mailbox names addressed in this Store, but is not considered part of them; this is important for Patterns in the Channels section. Note that you must append a slash if you want to specify an entire directory. (Default: none)

MaxSize size[k|m][b]

Messages larger than that will not be propagated into this Store. This is useful for weeding out messages with large attachments. K and M can be appended to the size to specify KiBytes resp. MeBytes instead of bytes. B is accepted but superfluous. If size is 0, the maximum message size is unlimited. (Default: 0)

MapInbox mailbox

Create a virtual mailbox (relative to Path) which aliases the INBOX. Makes sense in conjunction with Patterns in the Channels section, though with a Maildir slave, you probably want to place Inbox under Path instead. This virtual mailbox does not support subfolders.

Flatten delim

Flatten the hierarchy within this Store by substituting the canonical hierarchy delimiter / with delim. This can be useful when the MUA used to access the Store provides suboptimal handling of hierarchical mailboxes, as is the case with Mutt. A common choice for the delimiter is ..
Note that flattened sub-folders of the INBOX always end up under Path, including the "INBOXdelim" prefix.

Trash mailbox

Specifies a mailbox (relative to Path) to copy deleted messages to prior to expunging. See RECOMMENDATIONS and INHERENT PROBLEMS below. (Default: none)

TrashNewOnly yes|no

When trashing, copy only not yet propagated messages. This makes sense if the remote Store has a Trash as well (with TrashNewOnly no). (Default: no)

TrashRemoteNew yes|no

When expunging the remote Store, copy not yet propagated messages to this Store’s Trash. When using this, the remote Store does not need an own Trash at all, yet all messages are archived. (Default: no)

Maildir Stores
The reference point for relative Paths is the current working directory.

As mbsync needs UIDs, but no standardized UID storage scheme exists for Maildir, mbsync supports two schemes, each with its pros and cons.
The native scheme is stolen from the latest Maildir patches to c-client and is therefore compatible with pine. The UID validity is stored in a file named .uidvalidity; the UIDs are encoded in the file names of the messages.
The alternative scheme is based on the UID mapping used by isync versions 0.8 and 0.9.x. The invariant parts of the file names of the messages are used as keys into a Berkeley database named .isyncuidmap.db, which holds the UID validity as well.
The native scheme is faster, more space efficient, endianness independent and "human readable", but will be disrupted if a message is copied from another mailbox without getting a new file name; this would result in duplicated UIDs sooner or later, which in turn results in a UID validity change, making synchronization fail. The alternative scheme would fail if a MUA changed a message’s file name in a part mbsync considers invariant; this would be interpreted as a message deletion and a new message, resulting in unnecessary traffic.
is known to work fine with both schemes.
Use mdconvert to convert mailboxes from one scheme to the other.

Define the Maildir Store name, opening a section for its parameters.

AltMap yes|no

Use the alternative UID storage scheme for mailboxes in this Store. This does not affect mailboxes that do already have a UID storage scheme; use mdconvert to change it. See RECOMMENDATIONS below. (Default: no)

Inbox path

The location of the INBOX. This is not relative to Path, but it is allowed to place the INBOX inside the Path. (Default: ~/Maildir)

InfoDelimiter delim

The character used to delimit the info field from a message’s basename. The Maildir standard defines this to be the colon, but this is incompatible with DOS/Windows file systems. (Default: the value of FieldDelimiter)

SubFolders Verbatim|Maildir++|Legacy

The on-disk folder naming style used for hierarchical mailboxes. This has option has no effect when Flatten is used.
Suppose mailboxes with the canonical paths top/sub/subsub and INBOX/sub/subsub, the styles will yield the following on-disk paths:
- Path/top/sub/subsub and Inbox/sub/subsub (this is the style you probably want to use)
- Inbox/.top.sub.subsub and Inbox/..sub.subsub (this style is compatible with Courier and Dovecot - but note that the mailbox metadata format is not compatible). Note that attempts to set Path are rejected in this mode.
- Path/top/.sub/.subsub and Inbox/.sub/.subsub (this is mbsync’s historical style)
(Default: unset; will error out when sub-folders are encountered)

IMAP4 Accounts

Define the IMAP4 Account name, opening a section for its parameters.

Host host

Specify the DNS name or IP address of the IMAP server.
If Tunnel is used, this setting is needed only if SSLType is not None and CertificateFile is not used, in which case the host name is used for certificate subject verification.

Port port

Specify the TCP port number of the IMAP server. (Default: 143 for IMAP, 993 for IMAPS)
If Tunnel is used, this setting is ignored.

Timeout timeout

Specify the connect and data timeout for the IMAP server in seconds. Zero means unlimited. (Default: 20)

User username

Specify the login name on the IMAP server.

Pass password

Specify the password for username on the IMAP server. Note that this option is not required. If neither a password nor a password command is specified in the configuration file, mbsync will prompt you for a password.

PassCmd [+]command

Specify a shell command to obtain a password rather than specifying a password directly. This allows you to use password files and agents. The command must produce exactly one line on stdout; the trailing newline is optional. Prepend + to the command to indicate that it produces TTY output (e.g., a decryption password prompt); failure to do so will merely produce messier output.

Tunnel command

Specify a command to run to establish a connection rather than opening a TCP socket. This allows you to run an IMAP session over an SSH tunnel, for example.

AuthMechs type ...

The list of acceptable authentication mechanisms. In addition to the mechanisms listed in the SASL registry (link below), the legacy IMAP LOGIN mechanism is known. The wildcard * represents all mechanisms that are deemed secure enough for the current SSLType setting. The actually used mechanism is the most secure choice from the intersection of this list, the list supplied by the server, and the installed SASL modules. (Default: *)


Select the connection security/encryption method:
- no security. This is the default when Tunnel is set, as tunnels are usually secure.
- security is established via the STARTTLS extension after connecting the regular IMAP port 143. Most servers support this, so it is the default (unless a tunnel is used).
- security is established by starting SSL/TLS negotiation right after connecting the secure IMAP port 993.

SSLVersions [SSLv3] [TLSv1] [TLSv1.1] [TLSv1.2]

Select the acceptable SSL/TLS versions. Use old versions only when the server has problems with newer ones. (Default: [TLSv1] [TLSv1.1] [TLSv1.2]).

SystemCertificates yes|no

Whether the system’s default root cerificate store should be loaded. (Default: yes)

CertificateFile path

File containing additional X.509 certificates used to verify server identities. Directly matched peer certificates are always trusted, regardless of validity.
Note that the system’s default certificate store is always used (unless SystemCertificates is disabled) and should not be specified here.

ClientCertificate path

File containing a client certificate to send to the server. ClientKey should also be specified.
Note that client certificate verification is usually not required, so it is unlikely that you need this option.

ClientKey path

File containing the private key corresponding to ClientCertificate.

PipelineDepth depth

Maximum number of IMAP commands which can be simultaneously in flight. Setting this to 1 disables pipelining. This is mostly a debugging option, but may also be used to limit average bandwidth consumption (GMail may require this if you have a very fast connection), or to spare flaky servers like M$ Exchange. (Default: unlimited)

DisableExtension[s] extension ...

Disable the use of specific IMAP extensions. This can be used to work around bugs in servers (and possibly mbsync itself). (Default: empty)

IMAP Stores
The reference point for relative Paths is whatever the server likes it to be; probably the user’s $HOME or $HOME/Mail on that server. The location of INBOX is up to the server as well and is usually irrelevant.

Define the IMAP4 Store name, opening a section for its parameters.

Account account

Specify which IMAP4 Account to use. Instead of defining an Account and referencing it here, it is also possible to specify all the Account options directly in the Store’s section - this makes sense if an Account is used for one Store only anyway.

UseNamespace yes|no

Selects whether the server’s first "personal" NAMESPACE should be prefixed to mailbox names. Disabling this makes sense for some broken IMAP servers. This option is meaningless if a Path was specified. (Default: yes)

PathDelimiter delim

Specify the server’s hierarchy delimiter. (Default: taken from the server’s first "personal" NAMESPACE)
Do not abuse this to re-interpret the hierarchy. Use Flatten instead.


Define the Channel name, opening a section for its parameters.

{Master|Slave} :store:[mailbox]

Specify the Master resp. Slave Store to be connected by this Channel. If Patterns are specified, mailbox is interpreted as a prefix which is not matched against the patterns, and which is not affected by mailbox list overrides. Otherwise, if mailbox is omitted, INBOX is assumed.

Pattern[s] [!]pattern ...

Instead of synchronizing only one mailbox pair, synchronize all mailboxes that match the pattern(s). The mailbox names are the same on both Master and Slave. Patterns are IMAP4 patterns, i.e., * matches anything and % matches anything up to the next hierarchy delimiter. Prepending ! to a pattern makes it an exclusion. Multiple patterns can be specified (either by supplying multiple arguments or by using Pattern multiple times); later matches take precedence.
Note that INBOX is not matched by wildcards, unless it lives under Path.
The mailbox list selected by Patterns can be overridden by a mailbox list in a channel reference (a Group specification or the command line).
Example: "Patterns % !Trash"

MaxSize size[k|m][b]

Analogous to the homonymous option in the Stores section, but applies equally to Master and Slave. Note that this actually modifies the Stores, so take care not to provide conflicting settings if you use the Stores in multiple Channels.

MaxMessages count

Sets the maximum number of messages to keep in each Slave mailbox. This is useful for mailboxes where you keep a complete archive on the server, but want to mirror only the last messages (for instance, for mailing lists). The messages that were the first to arrive in the mailbox (independently of the actual date of the message) will be deleted first. Messages that are flagged (marked as important) and (by default) unread messages will not be automatically deleted. If count is 0, the maximum number of messages is unlimited (Default: 0).

ExpireUnread yes|no

Selects whether unread messages should be affected by MaxMessages. Normally, unread messages are considered important and thus never expired. This ensures that you never miss new messages even after an extended absence. However, if your archive contains large amounts of unread messages by design, treating them as important would practically defeat MaxMessages. In this case you need to enable this option. (Default: no).

Sync {None|[Pull] [Push] [New] [ReNew] [Delete] [Flags]|All}

Select the synchronization operation(s) to perform:
- propagate changes from Master to Slave.
- propagate changes from Slave to Master.
- propagate newly appeared messages.
- previously refused messages are re-evaluated for propagation. Useful after flagging affected messages in the source Store or enlarging MaxSize in the destination Store.
- propagate message deletions. This applies only to messages that are actually gone, i.e., were expunged. The affected messages in the remote Store are marked as deleted only, i.e., they won’t be really deleted until that Store is expunged.
- propagate flag changes. Note that Deleted/Trashed is a flag as well; this is particularly interesting if you use mutt with the maildir_trash option.
(--full on the command line) - all of the above. This is the global default.
(--noop on the command line) - don’t propagate anything. Useful if you want to expunge only.

Pull and Push are direction flags, while New, ReNew, Delete and Flags are type flags. The two flag classes make up a two-dimensional matrix (a table). Its cells are the individual actions to perform. There are two styles of asserting the cells:
In the first style, the flags select entire rows/colums in the matrix. Only the cells which are selected both horizontally and vertically are asserted. Specifying no flags from a class is like specifying all flags from this class. For example, "Sync Pull New Flags" will propagate new messages and flag changes from the Master to the Slave, "Sync New Delete" will propagate message arrivals and deletions both ways, and "Sync Push" will propagate all changes from the Slave to the Master.
In the second style, direction flags are concatenated with type flags; every compound flag immediately asserts a cell in the matrix. In addition to at least one compound flag, the individual flags can be used as well, but as opposed to the first style, they immediately assert all cells in their respective row/column. For example, "Sync PullNew PullDelete Push" will propagate message arrivals and deletions from the Master to the Slave and any changes from the Slave to the Master. Note that it is not allowed to assert a cell in two ways, e.g. "Sync PullNew Pull" and "Sync PullNew Delete Push" induce error messages.

Create {None|Master|Slave|Both}

Automatically create missing mailboxes [on the Master/Slave]. Otherwise print an error message and skip that mailbox pair if a mailbox and the corresponding sync state does not exist. (Global default: None)

Remove {None|Master|Slave|Both}

Propagate mailbox deletions [to the Master/Slave]. Otherwise print an error message and skip that mailbox pair if a mailbox does not exist but the corresponding sync state does.
For MailDir mailboxes it is sufficient to delete the cur/ subdirectory to mark them as deleted. This ensures compatibility with SyncState *.
Note that for safety, non-empty mailboxes are never deleted.
(Global default: None)

Expunge {None|Master|Slave|Both}

Permanently remove all messages [on the Master/Slave] marked for deletion. See RECOMMENDATIONS below. (Global default: None)

CopyArrivalDate {yes|no}

Selects whether their arrival time should be propagated together with the messages. Enabling this makes sense in order to keep the time stamp based message sorting intact. Note that IMAP does not guarantee that the time stamp (termed internal date) is actually the arrival time, but it is usually close enough. (Default: no)

Sync, Create, Remove, Expunge, MaxMessages, and CopyArrivalDate can be used before any section for a global effect. The global settings are overridden by Channel-specific options, which in turn are overridden by command line switches.

Set the location of this Channel’s synchronization state files. * means that the state should be saved in a file named .mbsyncstate in the Slave mailbox itself; this has the advantage that you needn’t to care for the state file if you delete the mailbox, but it works only with Maildir mailboxes, obviously. Otherwise this is interpreted as a string to prepend to the Slave mailbox name to make up a complete path.
This option can be used outside any section for a global effect. In this case the appended string is made up according to the pattern :master:master-box_:slave:slave-box (see also FieldDelimiter below).
(Global default: ~/.mbsync/).

name [channel[:box[,...]]] ...

Define the Group name, opening a section for its parameters. Note that even though Groups have an own namespace, they will "hide" Channels with the same name on the command line.
One or more Channels can be specified on the same line.
If you supply one or more boxes to a channel, they will be used instead of what is specified in the Channel’s Patterns. The same can be done on the command line, except that there newlines can be used as mailbox name separators as well.

Channel[s] channel[:box[,...]] ...

Add the specified channels to the group. This option can be specified multiple times within a Group.

Global Options
FSync yes

Selects whether mbsync performs forced flushing, which determines the level of data safety after system crashes and power outages. Disabling it is reasonably safe for file systems which are mounted with data=ordered mode. Enabling it is a wise choice for file systems mounted with data=writeback, in particular modern systems like ext4, btrfs and xfs. The performance impact on older file systems may be disproportionate. (Default: yes)

FieldDelimiter delim

The character to use to delimit fields in the string appended to a global SyncState. mbsync prefers to use the colon, but this is incompatible with DOS/Windows file systems. This option is meaningless for SyncState if the latter is *, obviously. However, it also determines the default of InfoDelimiter. (Global default: ; on Windows, : everywhere else)

BufferLimit size[k|m][b]

The per-Channel, per-direction instantaneous memory usage above which mbsync will refrain from using more memory. Note that this is no absolute limit, as even a single message can consume more memory than this. (Default: 10M)


If mbsync’s output is connected to a console, it will print progress counters by default. The output will look like this:

C: 1/2 B: 3/4 M: +13/13 *23/42 #0/0 S: +0/7 *0/0 #0/0

This represents the cumulative progress over channels, boxes, and messages affected on master and slave, respectively. The message counts represent added messages, messages with updated flags, and trashed messages, respectively. No attempt is made to calculate the totals in advance, so they grow over time as more information is gathered.


Make sure your IMAP server does not auto-expunge deleted messages - it is slow, and semantically somewhat questionable. Specifically, Gmail needs to be configured not to do it.

By default, mbsync will not delete any messages - deletions are propagated by marking the messages as deleted on the remote store. Once you have verified that your setup works, you will typically want to set Expunge to Both, so that deletions become effective.

mbsync’s built-in trash functionality relies on mbsync doing the expunging of deleted messages. This is the case when it propagates deletions of previously propagated messages, and the trash is on the target store (typically your IMAP server).
However, when you intend mbsync to trash messages which were not propagated yet, the MUA must mark the messages as deleted without expunging them (e.g., Mutt’s maildir_trash option). Note that most messages are propagated a long time before they are deleted, so this is a corner case you probably do not want to optimize for. This also implies that the TrashNewOnly and TrashRemoteNew options are typically not very useful.

If your server supports auto-trashing (as Gmail does), it is probably a good idea to rely on that instead of mbsync’s trash functionality. If you do that, and intend to synchronize the trash like other mailboxes, you should not use mbsync’s Trash option at all.

Use of the Trash option with M$ Exchange 2013 requires the use of DisableExtension MOVE due to a server bug.

When using the more efficient default UID mapping scheme, it is important that the MUA renames files when moving them between Maildir folders. Mutt always does that, while mu4e needs to be configured to do it:

(setq mu4e-change-filenames-when-moving t)


Changes done after mbsync has retrieved the message list will not be synchronised until the next time mbsync is invoked.

Using Trash on IMAP Stores without the UIDPLUS extension (notably, M$ Exchange up to at least 2010) bears a race condition: messages will be lost if they are marked as deleted after the message list was retrieved but before the mailbox is expunged. There is no risk as long as the IMAP mailbox is accessed by only one client (including mbsync) at a time.



Default configuration file


Directory containing synchronization state files


mdconvert(1), isync(1), mutt(1), maildir(5)

Up to date information on mbsync can be found at http://isync.sf.net/

SASL mechanisms are listed at http://www.iana.org/assignments/sasl-mechanisms/sasl-mechanisms.xhtml


Originally written by Michael R. Elkins, rewritten and currently maintained by Oswald Buddenhagen, contributions by Theodore Y. Ts’o.