Mattermost

mattermost.com
Mattermost

Self-hostable Slack alternative with native desktop, mobile and web apps, and many integrations. There's no end-to-end encryption, so the server can read all messages, and telemetry is on by default. The open source edition caps at 250 users.

Open Source

Mattermost Privacy Policy

Privacy Policy Summary

  • You are prohibited from sending chain letters, junk mail, spam or any unsolicited messages
  • Harming minors in any ways is prohibited
  • You are forced into binding arbitration in case of disputes
  • You are prohibited from posting content which promotes violence or politically or religiously extremist values
  • This service does not guarantee that it or the products obtained through it meet your expectations or requirements
  • The service is provided 'as is' and to be used at your sole risk
  • Invalidity of any portion of the Terms of Service does not entail invalidity of its remainder
  • If you are the target of a copyright claim, your content may be removed
  • Failure to enforce any provision of the Terms of Service does not constitute a waiver of such provision
  • The service provider makes no warranty regarding uninterrupted, timely, secure or error-free service
  • Specific content can be deleted without reason and may be removed without prior notice
  • You maintain ownership of your content
  • A complaint mechanism is provided for the handling of personal data
  • Extra data may be collected about you through promotions
  • This service is only available to users over 13 years old
  • Defend, indemnify, hold harmless; survives termination
  • Instructions are provided on how to submit a copyright claim
  • There is a date of the last update of the agreements
  • Information is provided about what kind of personal information is collected
  • Information is provided about how your personal data is used
  • You agree to indemnify and hold the service harmless in case of a claim related to your use of the service
  • You must report to the service any unauthorized use of your account or any breach of security
  • The service claims to be GDPR compliant for European users
  • Third parties used by the service are bound by confidentiality obligations
  • Your personal data may be used for marketing purposes
  • Your content can be deleted if you violate the terms
  • You agree not to use the service for illegal purposes
  • Information is provided about how your personal data is collected
  • First-party cookies are used
  • The service assumes no liability for any damages the user incurs
  • Information is gathered about you through third parties
  • This service gives your personal data to third parties involved in its operation
  • Your account can be deleted or permanently suspended without prior notice and without a reason
  • You agree not to submit libelous, harassing or threatening content
  • This service tracks which web page referred you to it
  • Third-party cookies are used for statistics
  • Location data may be collected, used and/or shared
  • Your profile is combined across various products
  • Your IP address is collected, which can be used to view your approximate location
  • Your data may be processed and stored anywhere in the world
  • Your personal data is used to employ targeted third-party advertising
  • Third-party cookies are employed, but with opt out instructions
  • Your personal information is used for many different purposes
  • Your personal data is aggregated into statistics
  • Your personal data is used for automated decision-making, profiling or AI training
  • Your biometric data is collected
  • The data retention period is kept to the minimum necessary for fulfilling its purposes
  • Your personal data may be sold unless you opt out
  • Tracking pixels are used in service-to-user communication
  • You can request access, correction and/or deletion of your data
  • This service shares your personal data with third parties that are not essential to its operation
  • The service claims to be CCPA compliant for California users
  • You can opt out of promotional communications
  • This service will continue using anonymized user-generated content after erasure of personal information
  • You can opt out of targeted advertising
  • Many different types of personal data are collected
  • You can retrieve an archive of your data
  • This service informs you that its Privacy Policy does not apply to third-party services
  • The court of law governing the terms is in a jurisdiction that is friendlier to user privacy protection (California, U.S.A.)
  • The court of law governing the terms is in the State of California, U.S.A.
  • Your personal data may be disclosed to comply with government requests without notice to you
  • Your personal data may be sold or otherwise transferred as part of a bankruptcy proceeding or other type of financial transaction

Score

E

Documents

About the Data

This data is kindly provided by tosdr.org. Read full report at: #2994

Mattermost Source Code

Author

mattermost

Description

Mattermost is an open source platform for secure collaboration across the entire software development lifecycle..

#collaboration#golang#hacktoberfest#mattermost#monorepo#react#react-native

Homepage

https://mattermost.com

Repository

  • LicenseOther
  • Created15 Jun 15
  • Primary languageTypeScript
  • Size1,396,319 KB
  • Stars39,582
  • Forks9,100
  • Watchers39,582

Language Usage

Language Usage

Project Health

  • Last commit5 minutes ago
  • Open issues1,109
  • Latest releasev12.1.0

Top Contributors

Recent Commits

  • cursor[bot](11 Nov 26)

    drip: AppError -> error: bot login check (#39580) Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: jesse <[email protected]>

  • Ben Schumacher(11 Nov 26)

    [MM-71680] Add Posts dead tuples health rule (#39732) * Add stale Posts vacuum health rule Co-authored-by: ben.schumacher <[email protected]> * Use registered mmctl health command in test Co-authored-by: ben.schumacher <[email protected]> * Drop pool/stats error split from support packet store diagnostics The real store never returns nil diagnostics, and partial failures already fail only the database_stats section on master. * Verify manual vacuum diagnostics timestamp Co-authored-by: ben.schumacher <[email protected]> * Fire DB_POSTS_VACUUM_STALE when Posts has no recorded vacuum Null last_vacuum and last_autovacuum mean no vacuum since the last statistics reset, which also resets n_dead_tup, so high dead tuples with no recorded vacuum is a finding rather than an unknown. * Handle active Posts vacuum in health rule Co-authored-by: ben.schumacher <[email protected]> * Avoid assuming Posts vacuum is idle Co-authored-by: ben.schumacher <[email protected]> * Detect VACUUM FULL on Posts and test vacuum-in-progress detection VACUUM FULL and CLUSTER report to pg_stat_progress_cluster and hold AccessExclusiveLock, so the Posts vacuum-in-progress check missed them. Check both progress views, each matched to its own lock mode. Neither view exposes relid or command to other roles without pg_read_all_stats, so the match still goes through the granted relation lock. Add a database-backed test that holds a VACUUM (FREEZE) of Posts open by pinning a heap page with a cursor, and asserts the flag is true during the vacuum, false after it finishes, and false while only an unrelated table is being vacuumed. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1s9qTtV8NkbQKPYzE * Fire on Posts dead tuples even after a recent vacuum The rule only fired when no vacuum had finished in seven days. Its 20% dead-tuple threshold matches autovacuum's default trigger, so a healthy autovacuum starts within about a minute of crossing it, and dead tuples that stay above it point to a problem either way. The seven-day gate resolved the rule in the common case where vacuum keeps running but cannot remove dead tuples because a long-running transaction, an inactive replication slot or a prepared transaction holds back cleanup. Fire whenever both thresholds are exceeded and no vacuum is running, and use the last vacuum time only to pick the message: a recent vacuum means cleanup is blocked, an old or missing one means autovacuum is not keeping up. The remediation now covers both causes. Rename the rule from DB_POSTS_VACUUM_STALE to DB_POSTS_DEAD_TUPLES, since it no longer requires a stale vacuum. The rule has not shipped yet. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1s9qTtV8NkbQKPYzE * Use n_live_tup for the Posts dead tuple ratio and harden vacuum test cleanup Compute the dead tuple ratio against n_live_tup from the same pg_stat_user_tables row as n_dead_tup instead of the support packet's exact post count. Both values are now sampled together, the estimate matches the kind autovacuum uses, and the rule no longer depends on the Stats section, whose COUNT(*) of Posts is slowest on the largest installs. The finding reports the estimate as live rows. Register the vacuum test's transaction release right after it begins, so a failed assertion before the vacuum starts ends the transaction instead of leaving its lock to block the DROP TABLE cleanup. Describe posts_vacuum_in_progress as covering VACUUM, VACUUM FULL and CLUSTER in the support packet comment. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1s9qTtV8NkbQKPYzE * Fix flaky Posts vacuum-in-progress test The test pinned a heap page and relied on VACUUM (FREEZE) waiting for a cleanup lock on it. The vacuum only waits when it has to freeze the pinned row, which needs the row's xmin to precede every snapshot in the database. On a pooled test store, a snapshot left open by an earlier test lets the vacuum skip the page and finish at once, so it never shows up in pg_stat_progress_vacuum and both subtests time out. Run the test in a fresh database with a minimal posts table and call the pg_stat_user_tables collector directly, so no other test shares the xmin horizon. Before starting the vacuum, wait for any older snapshot, such as an autovacuum worker's ANALYZE, to end, and name the blocking backends if one does not. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx3T9kUpDJqwe3XrSQP * Use pg_class.reltuples as the Posts row estimate n_live_tup is reset to zero by pg_stat_reset(), crash recovery and failover, and afterwards only counts inserts minus deletes until the next ANALYZE. On a large Posts table that inflates the dead tuple ratio and makes the rule fire while autovacuum is correctly idle. reltuples lives in pg_class, survives statistics resets, and is the estimate autovacuum itself compares n_dead_tup against. Rename posts_live_tuples to posts_estimated_rows accordingly. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1s9qTtV8NkbQKPYzE * Wait for older transactions in every database in the vacuum test The vacuum-in-progress test relies on VACUUM (FREEZE) waiting for a cleanup lock on a pinned page, which only happens when the probe row's xmin precedes the cursor's snapshot. A transaction still running in any database on the server holds back that snapshot's xmin, so on CI, where other test packages share the PostgreSQL server, the vacuum could skip the page and finish before it was observed. Wait for such transactions regardless of database; older snapshots still only matter in the test's own database. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1s9qTtV8NkbQKPYzE --------- Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: mattermost-code <[email protected]> Co-authored-by: maria.nunez <[email protected]> Co-authored-by: Claude <[email protected]>

  • Doug Lauder(11 Nov 26)

    MM-71792: Use dedicated database names in the shared channel test tool (#39901) * Use dedicated database names in the shared channel test tool Managed mode drops and recreates its two databases on every run. It used mattermost_test and mattermost_node_test, which are the databases that make run-server and make run-node use, so running the tool wiped a developer's local data. Managed mode now defaults to sharedchannel_test_a and sharedchannel_test_b. Unmanaged mode keeps the old names as defaults, since it connects to servers the developer started. New --db-a and --db-b flags override either default. * MM-71792: Validate database names in the shared channel test tool Custom --db-a/--db-b names went unquoted into DROP/CREATE DATABASE and the datasource URL. A hyphenated name broke the reset, and a mixed-case name made the reset target the lower-cased database while the server connected to the original one. Names are now restricted to lower case letters, digits and underscores. Managed mode refuses mattermost_test and mattermost_node_test, A and B must differ, and the flag help says the databases are dropped. * MM-71792: Add unit test for shared channel test tool DB name validation Covers valid names, the 63 character limit, invalid formats, the managed-mode guard on the make run-server/run-node databases, and duplicate names. * MM-71792: Quote database names in the shared channel test tool reset A validated name can still be a reserved word such as select, which breaks an unquoted DROP/CREATE DATABASE. Quote the identifier.

  • cursor[bot](11 Nov 26)

    Fix flaky TestUploadLicenseFile/try_to_get_one_through_trial,_with_TE_build (#39780) Automatic Merge

  • cursor[bot](11 Nov 26)

    [MM-71421] Keep typed minus sign in System Console number settings (#39907) An incomplete number such as "-" parses to NaN, which was rendered back into the controlled number input. React then overwrote the field, dropping the minus sign so typing "-1" saved 1. Render an empty value for NaN instead so the browser keeps the in-progress text. Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: mattermost-code <[email protected]>

  • Edgar Bellot Micó(11 Nov 26)

    MM-69954: Update websocket event handling for guest sessions (#37909)

  • Wayne Wollesen(11 Nov 26)

    Document the SCRAM-SHA-256 requirement for FIPS PostgreSQL auth (#39909) * Document the SCRAM-SHA-256 requirement for FIPS PostgreSQL auth The FIPS migration guide covered the 14-character Postgres password minimum but not that MD5 password authentication doesn't work with the FIPS image. Expand the prerequisite into steps that check how the password is stored, switch password_encryption to scram-sha-256, re-hash the password, and only then move pg_hba.conf entries off md5, so a still-running standard server isn't locked out mid-migration. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Address review: password timing, \password, scoped pg_hba entry Re-hashing an unchanged password is safe while Mattermost runs, but a new password breaks new connections from the running server, so that case moves to the start of the downtime window. Use psql's \password so the plaintext stays out of psql history and server logs, and scope the pg_hba.conf example to the Mattermost database and user so other users with MD5-stored passwords aren't locked out by a shared entry. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Address review: always check pg_hba, avoid a fixed container IP A SCRAM-stored password can still pass step a while the matching pg_hba.conf entry uses the cleartext password method, so a passing check now skips only the re-hash step; steps b and d always run. Swap the /32 example address for a placeholder and explain that recreated containers and pods can get a new IP, so the entry should cover their network range instead. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Set password_encryption with ALTER SYSTEM and recheck before re-hashing A password_encryption = 'md5' in postgresql.auto.conf overrides an edit to postgresql.conf, so the documented fix could leave MD5 in effect and \password would store another MD5 hash. Use ALTER SYSTEM, which writes to postgresql.auto.conf, and re-run SHOW password_encryption in the session used for the re-hash before continuing. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Handle password_encryption overrides and connection poolers Role and database settings override ALTER SYSTEM, so step b now lists them with \drds and resets any md5 entry, and step c runs SET password_encryption in the \password session so the re-hash is SCRAM regardless. Add a note that a pooler such as PgBouncer needs SCRAM on both connections: its own address and updated stored credentials toward PostgreSQL, and SCRAM for Mattermost's login to the pooler. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Address review: reload PgBouncer and replace MD5 client hashes PgBouncer keeps using old auth_file credentials until it's reloaded, and pg_reload_conf() doesn't reload it. Switching auth_type to scram-sha-256 also doesn't convert MD5 hashes already in auth_file, so the Mattermost entry needs a SCRAM secret or plaintext password. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Trim the pooler note to requirements and link PgBouncer's docs The step-by-step PgBouncer instructions were incomplete in ways that mattered (a newly generated auth_file SCRAM secret can't log in to PostgreSQL; auth_type is pooler-wide). State only what the migration needs: SCRAM on both pooler connections, the pooler's address in pg_hba.conf, and updating its stored credentials alongside step c. Link to PgBouncer's authentication settings for the how. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Replace the pooler note with a caution before the re-hash Pooler setup is outside this guide's scope, and each specific pooler statement had another PgBouncer configuration it didn't fit. Keep one caution, placed before step c because re-hashing can affect a pooler even while Mattermost is still running, and link to PgBouncer's authentication settings. Co-Authored-By: Claude Opus 5.5 <[email protected]> --------- Co-authored-by: Claude Opus 5.5 <[email protected]>

  • Wayne Wollesen(11 Nov 26)

    Use scram-sha-256 in the pg_hba.conf example for new installs (#39910) * Use scram-sha-256 in the pg_hba.conf example for new installs The database setup guide told admins to add an md5 pg_hba.conf entry. PostgreSQL v14+ already stores new passwords as SCRAM hashes, so the md5 entry negotiated SCRAM anyway, but it leaves an MD5 fallback that breaks FIPS-compliant images. Default the example to scram-sha-256, scope it to the mattermost database and mmuser role created earlier on the page, and warn that converting an existing md5 entry requires the stored password to be a SCRAM hash first. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Address review: scope the access claim, check password_encryption Don't claim the pg_hba.conf entry restricts all connections, since earlier or broader entries can still match; tell admins to review them. Note in the create-user step that SCRAM storage depends on password_encryption, with the fix if it's md5. Replace the link to the FIPS guide's conversion steps (not yet on master) with the check query and \password re-hash inline, so the page stands on its own. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Address review: set password_encryption with ALTER SYSTEM and recheck A password_encryption = 'md5' in postgresql.auto.conf overrides an edit to postgresql.conf, so the documented fix could leave MD5 in effect and the re-set password would be stored as MD5 again. Use ALTER SYSTEM, which writes to postgresql.auto.conf, and re-run SHOW password_encryption in the same session before resetting the password. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Address review: SET password_encryption per session, warn on poolers ALTER SYSTEM doesn't override per-role or per-database settings of password_encryption, so SHOW could stay md5. SET in the session that sets the password overrides every level, so use it in both the create-user step and the conversion step. Also warn that a pooler such as PgBouncer that authenticates with a stored MD5 hash can't connect once the password is re-hashed as SCRAM. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Address review: use the pooler's address in pg_hba.conf when present When PgBouncer or another pooler runs on its own host, PostgreSQL sees backend connections from the pooler, so an entry matching only the Mattermost server's address would reject them. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Address review: consolidate pooler guidance into one note Cover both pooler hops in a single note: the pooler-to-PostgreSQL connection (pooler's address in pg_hba.conf, stored credentials after re-hashing) and the Mattermost-to-pooler connection, which also needs SCRAM because FIPS images can't authenticate with MD5. Link to the PgBouncer configuration docs for the details. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Match the FIPS guide's PgBouncer reload and client-hash details Bring the pooler note in line with #39909: reload PgBouncer after updating its stored credentials (pg_reload_conf() doesn't reload it), and replace MD5 hashes in auth_file with a SCRAM secret or plaintext password, since changing auth_type doesn't convert them. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Trim the pooler note to requirements and link PgBouncer's docs The step-by-step PgBouncer instructions were incomplete in ways that mattered (a newly generated auth_file SCRAM secret can't log in to PostgreSQL; auth_type is pooler-wide). State only what Mattermost needs: SCRAM on both pooler connections, the pooler's address in pg_hba.conf, and a credential review when converting the password. Link to PgBouncer's authentication settings for the how. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Replace the pooler note with a caution before the re-hash Pooler setup is outside this page's scope, and each specific pooler statement had another PgBouncer configuration it didn't fit (most recently auth_user lookups). Keep one caution before the password conversion and link to PgBouncer's authentication settings. Co-Authored-By: Claude Opus 5.5 <[email protected]> --------- Co-authored-by: Claude Opus 5.5 <[email protected]>

  • cursor[bot](11 Nov 26)

    drip: go idioms: errors.AsType in app/webhook (#39912) Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: jesse <[email protected]>

  • Ben Schumacher(11 Nov 26)

    [MM-71814] Link plugins to their server logs (#39694) * Filter server logs by plugin Admins debugging a plugin had to know its ID and type it into the search, which also matched unrelated lines such as URLs containing the ID. The logs query now takes field filters (`fields`), which keep entries whose top-level field equals one of the given values. The Server Logs viewer uses them for a new Plugin menu that lists installed plugins by name, plus any plugin IDs seen in the loaded logs, so admins who can read logs but not plugins can still filter. The selection is kept in the URL as field.plugin_id parameters. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Link plugins to their server logs Plugin Management now links each plugin to its entries in the Server Logs viewer. A plugin that failed to start or kept crashing also links to only its errors and warnings, next to the notice that tells admins to check the logs. The links are left out for admins who can't open the Server Logs page. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Add a View logs button to plugin settings pages Checking a plugin's logs is a common task, so each plugin's settings page gets a View logs button next to the enable switch rather than an entry in the actions menu. It stays available to admins who can view the logs but can't change the plugin. When the plugin failed to start or keeps crashing, the failure notice also gets a button to its errors and warnings. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Show "All logs" when no plugin is picked in Server Logs "All plugins" suggested that only plugin entries were shown, though no plugin filter keeps the server's own entries too. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Drop the errors-only links from plugin failure notices The View logs button and link sit right next to the failure notices, so a second control that only preselected errors and warnings added noise. It also hid the info lines before a failure, which often explain it. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Address review findings on the Server Logs plugin filter - Count a picked plugin as a filter in the empty state, and clear it with the other filters. - Only read field filters from the URL that the viewer shows a filter for, so a link can't filter the view in a way the admin can't see or clear. - Don't fetch the installed plugins while plugins are turned off, which only logged an error into the log being viewed. - Ctrl/Cmd-clicking the only picked plugin restores all logs, as in the Levels menu. - Cut an oversized field name in the validation error's details. - Describe the new strings for translators. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Simplify the Server Logs plugin filter - Keep the picked plugins as a plain list, built into the generic fields filter only for the query and the URL, and read the URL's field.plugin_id values directly. - Use the plugins the admin sidebar already loads on every System Console page instead of fetching them again. - Share the "values seen so far" logic between known nodes and plugin IDs seen in the logs. - Don't reload when "All logs" is picked with no plugin picked. - On the server, check the field value as well as its name in the raw line, so lines from other plugins are skipped without decoding, and use model.StringSet for the value sets. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Simplify the plugin pages' View logs links - Decide whether to show the links with one helper next to pluginLogsPath, from the same console access and config that hide the Server Logs page, instead of two different checks. Plugin Management already gets both as props, so its container is back as it was. - Build the link from the picked plugins rather than the generic fields. - Keep the Server Logs path private to the module. - Describe the View logs string for translators. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Document the Server Logs plugin filter Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Document the View logs links to the plugin filter Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Log plugin health checks under plugin_id The Server Logs plugin filter matches plugin_id, so the health check lines that explain a crashing plugin were hidden from it. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Ask before View logs leaves unsaved plugin settings Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Open View logs on every level and node, and hide it for plain logs A URL that left out the levels or node fell back to the saved prefs, so a View logs link could open errors-only or on another node than the one that failed. The link now sets levels=all and an empty node, which mean every level and every node. Plain text logs have no plugin_id to filter on, so the link is hidden when LogSettings.FileJson is off. The admin definition now hides the Server Logs page with canViewServerLogs, so the links and the page can't drift apart. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Keep the levels and node in every Server Logs URL The viewer left all levels and all nodes out of the URL, so a URL copied after following a View logs link opened with the saved prefs of whoever opened it. The viewer now always writes levels and node. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Expect the empty node parameter in the shared server logs link e2e test The viewer now always writes the levels and node to the URL, so all nodes shows up as node= after the duration changes. --------- Co-authored-by: Claude <[email protected]>

  • cursor[bot](11 Nov 26)

    MM-71227: Update login session handling (#39251) * MM-71227: Update login session handling Co-authored-by: mattermost-code <[email protected]> * Address PR feedback: 1 answered, 2 resolved, 0 declined * Reject bot accounts in DoLogin A desktop token or SSO code-exchange code minted before ConvertUserToBot could still be redeemed afterwards, creating a session for the bot. ConvertUserToBot revokes existing sessions but cannot revoke an unredeemed code, so check IsBot at session creation time. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx3fMixBbUWN1JdX84U --------- Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: mattermost-code <[email protected]> Co-authored-by: Mattermost Build <[email protected]> Co-authored-by: Claude <[email protected]>

  • Ben Schumacher(11 Nov 26)

    [MM-71770] Add a unique packet_id to Support Packet metadata (#39876) * [MM-71770] Add a unique packet_id to Support Packet metadata Packets had no identifier of their own, so tools processing them could not reliably tell whether two archives were the same packet. GeneratePacketMetadata now sets a random PacketID, covering Support Packets and plugin packets. The field is optional when validating so metadata from older servers still parses. * [MM-71770] Include packet_id in the metadata golden fixture

  • Amy Blais(11 Nov 26)

    Don't fail the changelog run when the model drops a release note (#39685) * Don't fail the changelog run when the model drops a release note The model leaves out one or two notes per run, a different few each time, and require_sources_attributed then refused to write anything. - Skip PRs with an opt-out label (Changelog/Not Needed, ...) even when they have a release-note block. - Prompt: every note must appear; one bullet may carry several markers. - Retry once with only the dropped notes, merged into their sections. - A note still dropped no longer fails the run: it is listed under "Not placed by the model" in the report (and in the PR body outside dry runs), and its PR is left out of changelog-sources so the next run tries it again. - Keep the hard failure only for a PR recorded as a source with no bullet. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvehpkTndBYUCaHmZ91 * Address review: filter the retry answer, skip Go Version bullets, publish the not-placed list - Keep only bullets for the requested notes from the retry answer, so a repeated bullet is not inserted twice. - Bullets under ### Go Version are replaced or removed, so their PRs are not recorded as included. - Add a body_changed output and have the update publish step edit the PR body when only the not-placed list changed. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvehpkTndBYUCaHmZ91 * Address review: prune Go Version notes by ID, harden the retry filter, guard the body-only update - Prune notes placed only under ### Go Version by note ID, so another note of the same PR does not keep them referenced. - Normalize the retry answer before filtering so * and + bullets are matched, judge each bullet on every marker in its block, and keep an unmarked bullet only as the parent of a wanted one. - Before the body-only PR edit, re-read the body and skip with a warning if it changed during the run. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvehpkTndBYUCaHmZ91 * Simplify the retry filter and drop the body re-read guard - keep_requested_bullets goes back to a line filter on normalized text (so * and + bullets are matched); continuation lines follow their bullet. - Remove the PR body re-read before the body-only update. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvehpkTndBYUCaHmZ91 * Address review: retry Go Version-only notes, survive a failed retry, tidy the retry answer - Prune notes that appear only under ### Go Version before working out which notes are missing, so they are retried. - A failed retry request keeps the first answer instead of stopping the run. - Drop sections with no bullet, and text before the first heading, from the retry answer before merging, so valid bullets are kept and no empty section is added. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvehpkTndBYUCaHmZ91 * Clean up generated changelog text: doc links, repeated notes, restated migrations, Docs labels - Unlink a docs page the model put on a bullet whose markers do not name one of the page's PRs. - Keep each note in one section only; the first placement wins. - Write no bullet for a note that only restates a migration already listed under Database Schema Changes (it names the migration by number or name); the PR is still recorded as included and reported. - Strip "Docs:"/"Docs page:" labels from note text, also inline, and from the model's output. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvehpkTndBYUCaHmZ91 * Save the model's raw answers in the changelog run artifact The logs do not carry the model's output, so a note that went missing could not be traced. Write model_output.md next to sections.json, with the note ID each source was sent as and each answer (first pass and retry) before any cleanup. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvehpkTndBYUCaHmZ91 * Keep the subrouter prefix on API endpoint lines and compare against master's merge base config-change-checker wrote "Added `DELETE /` (`deletePropertyFieldOptions`) API endpoint." into release notes, and reported endpoints that master already has. - Resolve a Handle() registration's subrouter prefix from the PathPrefix(...).Subrouter() lines of api.go, and shorten {name:regexp} path variables to {name}. - Compare against the merge base with the current default branch instead of the parent of the PR's oldest commit, so work master took separately (the first commits of a stacked PR) is not counted as this PR's. - Flag API endpoint lines with no path in the changelog generator's note quality check. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvehpkTndBYUCaHmZ91 * Also compare config, API and audit changes against the PR's base branch tip A change counts as added only if it is in the head, not in the merge base, and not on the tip of the PR's own base branch (BASE_REF). It counts as removed only if it is in the merge base, not in the head, and still on that tip. A stacked branch that never took master in after its first PR was squashed no longer re-reports that PR's endpoints, and a cherry-pick to release-X.Y is compared with that release branch. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvehpkTndBYUCaHmZ91 * Make the changelog change smaller: drop the extra Docs stripping, narrow the migration rule - Remove the inline and model-output stripping of Docs:/Docs page: labels; the line form from #39380 stays. - A note is dropped as restating a migration only when it does nothing but name it ("Added migration 000238."); a note that says what the change is for is kept. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvehpkTndBYUCaHmZ91 * Keep a recorded PR's dropped note listed; shorten path variables with braced patterns - A note dropped from a PR whose other note was placed: the PR is recorded, so it isn't sent again, and the next run rebuilt the not-placed list without it. The list is now a task list, and such a note is carried over from the last PR body until a reviewer adds it by hand and ticks it. A note whose PR isn't recorded is still retried, as before. - A path variable whose pattern has braces of its own, /{fingerprint:[a-f0-9]{32}}/mute in api4/health_dashboard.go, became /{fingerprint}}/mute. The variable now ends at its matching brace. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvgCUhLXZqybdTPm243 * Don't carry a placed note on the not-placed list; compare release-branch PRs with the release line - A note listed while its PR wasn't recorded, then placed by a later run: the PR is now recorded, and the item was carried over as if it were still missing. Only a PR not sent this run is carried; one sent this run is either placed or listed afresh. - A PR to release-X.Y was compared with its merge base with master, so a route the release line added after the cut and the PR removes wasn't reported, and a file the line added looked new. Such a PR is now compared with its merge base with the release branch. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01CdqbvgCUhLXZqybdTPm243 --------- Co-authored-by: Claude <[email protected]>

  • Ben Schumacher(11 Nov 26)

    [MM-71813] Filter server logs by plugin (#39693) * Filter server logs by plugin Admins debugging a plugin had to know its ID and type it into the search, which also matched unrelated lines such as URLs containing the ID. The logs query now takes field filters (`fields`), which keep entries whose top-level field equals one of the given values. The Server Logs viewer uses them for a new Plugin menu that lists installed plugins by name, plus any plugin IDs seen in the loaded logs, so admins who can read logs but not plugins can still filter. The selection is kept in the URL as field.plugin_id parameters. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Show "All logs" when no plugin is picked in Server Logs "All plugins" suggested that only plugin entries were shown, though no plugin filter keeps the server's own entries too. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Address review findings on the Server Logs plugin filter - Count a picked plugin as a filter in the empty state, and clear it with the other filters. - Only read field filters from the URL that the viewer shows a filter for, so a link can't filter the view in a way the admin can't see or clear. - Don't fetch the installed plugins while plugins are turned off, which only logged an error into the log being viewed. - Ctrl/Cmd-clicking the only picked plugin restores all logs, as in the Levels menu. - Cut an oversized field name in the validation error's details. - Describe the new strings for translators. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Simplify the Server Logs plugin filter - Keep the picked plugins as a plain list, built into the generic fields filter only for the query and the URL, and read the URL's field.plugin_id values directly. - Use the plugins the admin sidebar already loads on every System Console page instead of fetching them again. - Share the "values seen so far" logic between known nodes and plugin IDs seen in the logs. - Don't reload when "All logs" is picked with no plugin picked. - On the server, check the field value as well as its name in the raw line, so lines from other plugins are skipped without decoding, and use model.StringSet for the value sets. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Document the Server Logs plugin filter Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB * Log plugin health checks under plugin_id The Server Logs plugin filter matches plugin_id, so the health check lines that explain a crashing plugin were hidden from it. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1pSsUR8DWGR89iosB --------- Co-authored-by: Claude <[email protected]>

  • Ben Schumacher(11 Nov 26)

    MM-40523: Follow server logs with mmctl logs -f (#39646) * MM-40523: Follow server logs with mmctl logs -f mmctl logs now reads the logs through POST /api/v4/logs/query, so they can be filtered on every node, and can follow them: - --follow/-f keeps polling every --interval (default 5s, at least 2s) and prints new entries until interrupted. Each poll reaches back 30s before the previous successful fetch, and entries already printed are dropped per node, with a bounded set of keys. - --level (error, warn, info, debug groups), --node, --search, --request-id and --since map to the query's filters, and are also applied on the client for servers that ignore them. - Unreachable nodes are reported once on stderr until they recover, and lines are prefixed with their node when several nodes are listed. - --plain keeps the previous output from GET /api/v4/logs. Plain-text log files fail with a message pointing to it, and servers without the query endpoint fall back to it when no query-only flag is set. The query endpoint is now also served over the local socket, so mmctl --local logs keeps working. newTestCmd now resets slice flags without adding their printed default as a value. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Remove an accidentally committed Redis dump Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Drop the --request-id flag from mmctl logs --search covers it: a request ID is matched like any other value. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Sort followed log entries with slices.SortStableFunc check-style's sortSlices rule doesn't allow sort.SliceStable. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Use the logrus format for mmctl logs by default - The log parser now reads mlog's "timestamp" field, so entries show when the server logged them. Before, logrus printed the time mmctl printed the line, and the default format only showed the timestamp as a field. - mmctl logs always formats with logrus, coloring the level when printing to a terminal. mmctl decides on the colors itself, as logrus can't see the terminal behind the node prefix. NO_COLOR, CLICOLOR and CLICOLOR_FORCE are honored. - The logrus writer strips escape sequences from the message, caller and fields, as LogEntry.String() did for the default format. - --logrus/-l is hidden and deprecated, and still accepted. - Added -s as a shorthand for --search. - Removed the now unused SimpleWriter. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Document the new mmctl logs options Covers following and filtering the logs, the JSON file logging they need, --plain, the colored output and NO_COLOR/CLICOLOR_FORCE, -s, and the deprecated --logrus flag. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Fix review findings in mmctl logs - Entries with mlog levels logrus doesn't have printed an extra "not a valid logrus Level" line and showed as info. critical now shows as error, stdlog and custom levels as info, and the server's level is kept as an original_level field. - The follow loop checks the context after each tick, as select picks either case when both are ready. This made TestLogFollower flaky and could send one more request after Ctrl-C. - Each queried entry is parsed once, by the human parser, and keyed on its raw bytes, which the server writes with sorted keys. The parser now keeps numbers as written. - newTestCmd gives each test fresh slice flag values and restores the originals, as pflag can't reset a slice value's first-Set state. - Removed LogEntry.String(), unused since SimpleWriter went away. Its control character and multi-line cases moved to the logrus writer tests. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Keep mmctl logs -f from printing entries before the first page The first poll reaches 30 seconds back, so entries a full first page left out were printed after newer ones. A node whose first response fills a page now gets a floor at the oldest entry printed, and later polls drop anything older. With --number 0, the first query fetches one entry per node without printing it, so the floor follows the server's clock and only new entries print. The e2e follow test now allows three polls for the marker, as the log file may only show it some time after Flush. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Describe mmctl logs --plain accurately --plain prints the logs unfiltered through GET /api/v4/logs, as before, which formats JSON lines rather than printing them raw. The help, errors, warning and docs now say so. The --since example now raises --number, as it otherwise prints at most 200 entries per node. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Drop "as earlier versions of mmctl did" from the --plain help Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Fall back to unfiltered logs instead of mmctl logs --plain When the query returns nothing and the server's log file turns out to be plain text, mmctl logs now prints the unfiltered log lines with a warning, as it already did when the server can't serve the query. With --follow, --level, --node, --search or --since, it still fails, as those need JSON file logging. That leaves --plain with nothing to do. It's hidden and deprecated, and still accepted for existing scripts. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Check that mmctl logs prints a full first page against a server A full first page sets each node's floor at its first entry, which relies on /logs/query returning each node's entries oldest first. The unit tests mock that order, so an e2e test now checks it against a real server. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Skip cluster separators when mmctl logs checks the log format On a cluster, GET /api/v4/logs puts a separator before each node's lines, so the first line it returns is always a rule. The plain-text check took it for a plain-text log line, so a query that matched nothing failed, or fell back to the unfiltered logs, on every cluster. The check now skips blank lines, rules and node names, and looks at the first log line. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Mark successful log queries as successful in the audit log getLogs and queryLogs created their audit record as failed and never called Success(), so every successful request was audited as a failure. With mmctl logs --follow and audit logging on, that's one failed queryLogs record per poll. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Address review feedback on mmctl logs - Drop --plain. It never shipped, so no script can rely on it. - Cap --number at 10000 with a warning instead of failing. The server has always capped the page there, and earlier versions of mmctl printed what it returned. - Print warnings through printer.PrintWarning, so they're sanitized and --quiet hides them. - Warn when a poll fills a page with entries no older than its start. The server fills the page from the newest entry back, so the older entries in the window never print. A full page that reaches past the start, as remote nodes on servers older than v12.2 send, doesn't warn. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx4o2N8Az1jpckmGYuo * Document that mmctl logs filtering needs a v12.2 server * Refuse mmctl logs filters and follow on servers older than v12.2 Servers older than v12.2 ignore --search, don't filter the logs of other nodes, and read their whole log file on every --follow poll. When --follow, --level, --search or --since is set, mmctl logs now pings the server, reads its version from X-Version-Id, and fails with a clear error when the server is older than v12.2. --node works on older servers and stays allowed. Without a version it can read, mmctl lets the query run. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1gAYPuLZjSp32iFU7 * Stop the warning loop when following plain-text logs queryLogs logged a warning for every log line that isn't JSON. With mmctl logs --follow on a log file that holds plain-text lines, each warning was another line for the next poll to read back, so the log file grew by about 2000 lines a second until it rotated. - queryLogs now logs one warning per node, with the number of lines it skipped, so the warnings can't feed the next query. - mmctl logs reads LogSettings.FileJson from the config, as the System Console does. With --follow, it fails before querying when the log file is plain text. When a query returns no entries, the config decides whether to fall back to unfiltered logs. Without the permission to read the config, mmctl checks the last log line as before. A query that returns entries without --follow doesn't read the config. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1gAYPuLZjSp32iFU7 * Only warn about skipped log entries when some were skipped A poll reaches 30s further back than the previous one started, so a full page whose entries were all no older than the poll's date_from could still overlap entries already printed. mmctl warned about skipped entries on every such poll, about 285-2000 entries a second per node at the default interval, with nothing lost. mmctl now keeps the newest timestamp it printed for each node, and only warns when the full page's oldest entry is newer than that. Both timestamps come from the node's clock, so clock skew between mmctl and the server doesn't affect it. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1gAYPuLZjSp32iFU7 * Log the log lines that aren't JSON once per scan GetLogsSkipSend logged "Failed to parse line, skipping" at debug level for every line it couldn't parse. With the file at debug level, each of those messages was another line for the next query of a client polling /logs/query to read back, so a plain-text log file grew on every poll. It now counts those lines and logs one debug message with the count at the end of the scan. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1gAYPuLZjSp32iFU7 * Refuse mmctl logs --node on servers older than v12.2 Servers older than v12.2 only return the local node's logs when it's named "default". On a cluster, --node with the name of the node mmctl connects to returned nothing for it, without saying why. --node now joins --follow, --level, --search and --since in the server version check. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1gAYPuLZjSp32iFU7 * Don't log the log lines queryLogs skips queryLogs logged a warning, once per node and request, for the log lines that aren't JSON. A log file holds them when LogSettings.FileJson is off, which an admin can't fix by reading the warning, and clients already check FileJson. At the default level, every poll of a live tail added a line to the log file, which was another line for the next poll to skip. The scan in GetLogsSkipSend still logs one debug message with the count. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1gAYPuLZjSp32iFU7 * Read the log format before every mmctl logs query After LogSettings.FileJson is turned off, the log file still holds the JSON lines from before until it rotates. The query skips the new plain-text lines and returns those old entries, so mmctl logs printed them as the latest logs, and never fell back to the unfiltered logs. Only --follow read the config before querying. mmctl logs now reads LogSettings.FileJson before the first query in every mode. When the log file is plain text, it prints the logs unfiltered, or fails when the flags need the query, without querying. Without the permission to read the config, an empty result still decides by the last log line. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx1gAYPuLZjSp32iFU7 --------- Co-authored-by: Claude <[email protected]>

  • Jesse Hallam(11 Nov 26)

    MM-70987: reject SHA-1 detached signatures via go-crypto openpgp/v2 (#39840) * MM-70987: verify detached signatures with go-crypto openpgp/v2 * MM-70987: reject signatures from keys revoked after signing; cover weak key rejection

  • Ben Schumacher(11 Nov 26)

    [MM-70315] Reload SAML SP when certificates are rotated (#38383) * MM-70315: Reload SAML SP when certificates are rotated Replacing an already-configured SAML certificate through the System Console reused the same on-disk filename, so UpdateConfig produced a byte-identical config, hasChanged was false, and the reload listener that runs ConfigureSP never fired. The in-memory Service Provider kept the stale certificate until a restart. Reconfigure the SP directly from the certificate write paths, decoupled from the config diff, and broadcast a new ClusterEventReloadSaml so peer nodes rebuild their SP too (the config-change message can't carry this change because the config is unchanged). Reload errors are logged, not returned, so uploading one certificate before the rest is not fatal. Co-Authored-By: Claude Opus 4.8 <[email protected]> * MM-70315: Unit-test reloadSaml wiring in the OSS server The end-to-end reload is covered in the enterprise repository, but the open-source wiring around it was untested. Cover reloadSaml: a nil SAML interface (team edition) is a no-op, a successful reconfigure broadcasts ClusterEventReloadSaml to peers, and a ConfigureSP error is swallowed without aborting the cluster broadcast. Co-Authored-By: Claude Opus 4.8 <[email protected]> * MM-70315: Consolidate SAML reload wiring Route every SP reconfigure (startup, config listener, cluster handler, reloadSaml) through a single Channels.configureSaml helper, and send the reload_saml cluster message via a.Cluster() like other app-level events instead of a SAML-specific PlatformService method. The four certificate write paths now share saveSamlCertificateConfig, so a new upload path can't forget the forced reload. SetSamlIdpCertificateFromMetadata now rejects data that doesn't decode as PEM with a 400 instead of dereferencing a nil block. Tests: table-drive the reloadSaml cases, share the mock SAML setup, and cover clusterReloadSamlHandler. * MM-70315: Test reload_saml registration and the metadata upload path Invoke the reload_saml handler through registerClusterHandlers instead of calling it directly, and assert the broadcast is reliable and non-waiting. Cover the nil-PEM guard in SetSamlIdpCertificateFromMetadata and a same-filename rotation through it end to end. * MM-70315: Share the SAML certificate upload path and trim test setup Fold the three AddSaml*Certificate bodies into addSamlCertificate, return 400 when metadata certificate data doesn't parse, and run the no-op and invalid-PEM cases on a bare App with an ECDSA test certificate. * MM-70315: Reject invalid SAML IdP certificates and save SAML settings cluster-wide Uploading a file that isn't a PEM certificate used to be accepted and only failed in ConfigureSP, which now keeps trusting the previous certificate. Both IdP certificate paths share setSamlIdpCertificate and reject such data with a 400. SAML certificate uploads and removals save the config with SaveConfig, so the other nodes receive the new filenames and a failed save is reported instead of logged. * MM-70315: Reload SAML even when the config save fails The certificate is already written by then, so skipping the reload left the node and its peers on the old certificate whenever SaveConfig failed, e.g. because a peer was unreachable. Pass a field selector instead of repeating the certificate filename, and share the reload test setup and assertions. * MM-70315: Show certificate upload errors in the System Console The file upload setting read its error from a hardcoded idpCertificateFileError state key that nothing wrote, and upload failures were never stored at all. Now that the server rejects invalid IdP certificates with a 400, the console silently ignored the rejection. Store upload errors under the setting's own key, matching the remove and metadata paths, and render from that key. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx3hv1YZB1KCMBKEsfB * MM-70315: Give rejected IdP certificate uploads their own error The upload path reused the metadata error, which told admins the "metadata information received from the Identity Provider" failed to parse even when they had uploaded a file. Split parsing out so each caller returns its own error ID, and add one for file uploads. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx3hv1YZB1KCMBKEsfB * MM-70315: Reject invalid SAML SP certificates and private keys A failed reload keeps the last good SP, so an SP certificate or private key that can't be loaded was accepted with a 200 and silently ignored until the next restart, when SAML would break. Reject such uploads with a 400, as the IdP certificate upload already does. Validate each file the way tls.X509KeyPair parses it, since that is how the SAML implementation loads the key pair, and check every test case against tls.X509KeyPair so the two can't drift. The certificate and key aren't checked against each other: during a rotation they're uploaded one at a time, so the first upload always mismatches the file on disk. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx5HmkYgCSaeevrt75t * MM-70315: Document the SAML certificate and key formats The settings reference didn't say what format the IdP certificate, SP certificate and SP private key files take. Uploads in other formats are now rejected, so state the PEM requirement and how to convert a DER certificate. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx5HmkYgCSaeevrt75t * MM-70315: Add an i18n string for an unconfigured SAML Service Provider Used by the enterprise SAML implementation when ConfigureSP has never succeeded, e.g. because a stored certificate or key failed to load at startup, so login, metadata and IdP responses report that SAML is not configured instead of a misleading certificate error. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx5C8B4VYiSZLS5Xtec --------- Co-authored-by: Claude Opus 4.8 <[email protected]>

  • cursor[bot](11 Nov 26)

    [MM-71420] Check MaxUsersPerTeam before bulk import instead of per line (#39868) * [MM-71420] Check MaxUsersPerTeam before bulk import instead of per line Bulk import enforced TeamSettings.MaxUsersPerTeam per user line. With concurrent workers the count-then-insert check raced, so teams could end up over the limit, and once a line did hit it the job aborted partway, leaving the rest of the file unimported. The limit is no longer enforced while memberships are written; a warning is logged when a team ends up past it. Instead, a new pre-flight pass counts the active members each team would gain from the file and fails the import before anything is written if a team would go over the limit. --skip-preflight downgrades that to a warning. Co-authored-by: mattermost-code <[email protected]> * Address PR feedback: 0 answered, 2 resolved, 0 declined --------- Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: mattermost-code <[email protected]>

  • cursor[bot](10 Nov 26)

    [MM-71780] Allow forwarding a post to an unjoined public channel (#39904) * [MM-71780] Allow forwarding a post to an unjoined public channel Enable the Forward button for public channels the user hasn't joined when they can join public channels on the team, and join the channel before forwarding the post instead of after behind an impossible condition. Co-authored-by: mattermost-code <[email protected]> * Add fallback message when joining the selected channel fails without an error message Co-authored-by: mattermost-code <[email protected]> --------- Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: mattermost-code <[email protected]>

  • cursor[bot](10 Nov 26)

    [MM-71767] Fix scheduled post inline edit keeping cancelled drafts and closing on save errors (#39880) Automatic Merge

  • Maria A Nunez(10 Nov 26)

    Bump @mattermost/compass-ui to 0.1.0-alpha.28 (#39896) Refresh snapshots for the new CSS module class hashes and expect AdminPanel header clicks to toggle expandable panels. Co-authored-by: Cursor Agent <[email protected]>

  • cursor[bot](10 Nov 26)

    MM-71543: Update search popout team handling (#39625) * MM-71543: Update search popout team handling Co-authored-by: mattermost-code <[email protected]> * Refactor: Export ID_PATH_REGEX from utils/path to avoid duplication - Export ID_PATH_REGEX from utils/path.ts - Update mention_search.ts to use imported regex instead of creating its own - Update rhs_search_popout.tsx to use imported regex instead of creating its own - Removes code duplication by centralizing regex pattern Co-authored-by: Devin Binnie <[email protected]> --------- Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: mattermost-code <[email protected]> Co-authored-by: Devin Binnie <[email protected]>

  • Jesse Hallam(10 Nov 26)

    Export Client4 request helpers for fully-qualified URLs (#39863)

  • jackbenningdev(10 Nov 26)

    MM-71667: Add outage catch-up and cursor rewind scenarios to the shared channel test tool (#39778) * MM-71667: Add outage catch-up and cursor rewind scenarios to the shared channel test tool The outage catch-up scenario stops Server B, changes older posts (edits, reactions, acknowledgements, a reply, a delete) behind enough new posts to span several sync batches, restarts Server B and compares both channels. The cursor rewind scenario writes content before sharing a channel, then a small test plugin moves the remote's cursor back with UpdateSharedChannelCursor, and all of the content must reach Server B. Managed servers now get their own temporary plugin directories so the test plugin stays out of ServerDir. * MM-71667: Address review feedback on the shared channel test tool Wait for the full timeout before reporting differences, record a failure when the rewind test is cancelled, bound each cursor request, and pin the plugin signature and sync batch settings on the managed servers. * MM-71667: Bound post reads and report the final read failure in the shared channel test tool Each poll gets a 30 second timeout, a read that still fails at the deadline is reported instead of earlier differences, and a failed Server B restart after a failed mutation is recorded. * MM-71667: Bound shared channel test tool reads by the polling deadline Retry the cursor plugin request only while its route is missing, time out the read before the rewind, and force post priority on so post reads include acknowledgements. --------- Co-authored-by: Jack Benning <[email protected]> Co-authored-by: wiggin77 <[email protected]>

  • cursor[bot](10 Nov 26)

    drip: go idioms: errors.AsType in app/command (#39870) Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: jesse <[email protected]>

  • Nick Misasi(10 Nov 26)

    MM-71544: Configurable cloud preview landing path and welcome modal fallback (#39862) Co-authored-by: Cursor Agent <[email protected]>

  • cursor[bot](10 Nov 26)

    [MM-70868] Return 404 instead of 403 when a post handler cannot find the post (#39861) * [MM-70868] Return 404 instead of 403 when a post handler cannot find the post getEditHistoryForPost, updatePost, postPatchChecks (patch and restore), saveIsPinnedPost (pin and unpin) and restorePostVersion replaced the GetSinglePost error with a permission error, so a missing post was reported as "You do not have the appropriate permissions." Pass the AppError through, as deletePost already does. Co-authored-by: mattermost-code <[email protected]> * [MM-70868] Document 404 responses for post update, patch, pin and unpin Co-authored-by: mattermost-code <[email protected]> * Retrigger CI and review Co-authored-by: mattermost-code <[email protected]> * [MM-70868] Fix test lint shadow and keep restore version lookup masked Co-authored-by: mattermost-code <[email protected]> --------- Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: mattermost-code <[email protected]>

  • cursor[bot](10 Nov 26)

    MM-71041, MM-71065: Update reference-style link handling in the shared markdown parser (#38900) * MM-71041: Update reference-style link resolution in the markdown parser Resolve reference labels through an index built once per document, and read a paragraph's reference definitions from a single joined copy of its text. Co-authored-by: mattermost-code <[email protected]> * MM-71041: Extend reference-link parsing coverage Add rows for container blocks, definitions continued onto a second line, label folding boundaries, caller-supplied definition sets, and concurrent resolution against one parse. Co-authored-by: mattermost-code <[email protected]> * MM-71041: Skip the wall-clock parse bound under -race and -short The concurrent parses still run; only the one-second assertion is skipped when the race detector is on or tests are run in short mode. Co-authored-by: mattermost-code <[email protected]> * Address PR feedback: 0 answered, 1 resolved, 0 declined * Address PR feedback: 1 answered, 7 resolved, 1 declined * Split reference-definition growth tests by ticket and drop redundant concurrency subtests * Replace reference definition slice with indexed ReferenceDefinitions type --------- Co-authored-by: Cursor Agent <[email protected]> Co-authored-by: mattermost-code <[email protected]> Co-authored-by: Mattermost Build <[email protected]> Co-authored-by: Jesse Hallam <[email protected]>

  • Marco(10 Nov 26)

    Refresh the NGINX reverse proxy documentation (#39776) * Refresh the NGINX reverse proxy documentation Make setup-nginx-proxy the single authoritative NGINX page and align the other pages with it. Pages that show NGINX snippets keep only their own options and link to it. The high availability page keeps a runnable script that writes a port 80 site file with a multi-node upstream. Pages and structure - Merge "Configure NGINX as a proxy for Mattermost server" and the separate SSL section into one "Configure NGINX with SSL and HTTP/2" section with a complete site file: HTTP to HTTPS redirect server and an HTTPS server with a TLS profile placeholder, HSTS and two locations. Inbound links and a stale redirect anchor are repointed. - TLS profiles adapted from Mozilla 6.0: Compatible (TLS 1.2 and 1.3) and Modern (TLS 1.3 only), plus an optional OCSP stapling note. - Certbot moved to its own section (the certificate source is the customer's choice): standalone for the first certificate, then webroot renewals with a deploy hook. - Setup steps: restart instead of reload (also starts a stopped NGINX), RHEL/CentOS firewalld and SELinux hints, TrustedProxyIPHeader set to X-Real-IP, and a verification step that checks status 200 and HSTS. - Scaling section: nginx.conf tuning, multi-node upstream, reuseport, profile image cache with a cache directory step. - FAQ entries rewritten: subpath "Too many redirects", WebSocket 403, Docker, GitLab and Certbot http-01. - Every directive line in config examples has a short trailing comment. Location blocks (WebSocket and /) - WebSocket: proxy_read_timeout 120s (above the 60s server ping), proxy_connect_timeout 10s. Removed send_timeout, proxy_send_timeout and proxy_buffer_size, which have no effect on upgraded connections. - location /: proxy_read_timeout 300s (Mattermost read and write timeout), proxy_connect_timeout 10s, proxy_buffer_size 16k with proxy_buffers 8 16k (16k alone fails nginx -t on 4K page hosts), client_max_body_size 101M (100 MiB MaxFileSize plus overhead). - Upstream: keepalive 32 and keepalive_timeout 55s, so NGINX closes idle connections before Mattermost's 60s idle timeout. Removed - The 0-RTT ssl_early_data and X-Early-Data block. - Defaults and redundant options: tcp_nodelay off, use epoll, conditional logging, X-Frame-Options as a request header and others. - The cache snippet files and the cache lines in the basic config. The cache now lives in the scaling section, with max_fails=3 fail_timeout=10s instead of max_fails=0. - The "Key optimizations" explanation lists, replaced by inline comments. Other pages - Pre-authentication secrets: partial config with the secret check in the WebSocket location and location /; HSTS repeated in the rejection because add_header in a location hides the server-level one. - SSL client certificates: client certificate and device ID headers per location, 2048 bit keys. - High availability: script updated to the main page state. - Deploy containers: inline example replaced by a link; Docker note and rate-limiting snippet dropped. - Environment settings: TrustedProxyIPHeader guidance and a warning against X-Forwarded-For. - Rate limiting: useBaseUrl for a raw link. - Transport encryption, TLS setup, session attributes, mobile troubleshooting: aligned with the main page. Known trade-off: the profile image cache ignores Mattermost's private Cache-Control header (proxy_ignore_headers), so NGINX can serve a cached avatar to anyone who requests the same URL, even without a login. The {26} in the location regex limits the cache to requests by user ID and excludes /api/v4/users/me/image. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> * Require NGINX real_ip behind a load balancer in rate limiting docs The VaryByHeader row recommended X-Real-IP for NGINX without a condition, and offered TrustedProxyIPHeader alone as an alternative behind a load balancer. In that setup X-Real-IP holds the load balancer's address, so all clients share one rate limit quota and can receive 429 responses. State that the NGINX real_ip module with set_real_ip_from, limited to the load balancer's addresses, must be configured first. Keep the warning against X-Forwarded-For. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> * Address review feedback on the NGINX documentation Remove the profile image cache - Drop the image location with proxy_ignore_headers and proxy_cache_valid from the main page and the high availability script. Mattermost marks profile images private and requires a session, so the cache could serve an avatar to anyone who requests the same URL. The static file cache in location / stays. - Remove the pointer to the image cache from the pre-authentication secrets page. Setup steps and settings - Run "nginx -t && systemctl restart nginx" on the main and high availability pages, so NGINX is only restarted with a valid config. - OCSP stapling: point ssl_trusted_certificate to a separate CA bundle with the root and intermediate certificates. - TLS profile: on NGINX before 1.29.2 ssl_protocols is taken from the default HTTPS server, so add default_server or use its own IP when other HTTPS sites share port 443. - Public port other than 443: keep the port in the HTTP to HTTPS redirect, because $host does not include the port. Certbot and troubleshooting - Replace the standalone fallback with "certbot reconfigure --authenticator standalone" scoped to the certificate with pre and post hooks. Hooks given to "certbot renew" are only saved after a successful renewal, and options apply to every certificate. - WebSocket 403 FAQ: the combined access log does not record Origin and Host, so add a temporary log format that does. Other pages - SSL client certificates: set the same headers in every other location that proxies to Mattermost. - Environment settings: keep Mattermost reachable only through the proxy, apart from machines that you fully trust. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> --------- Co-authored-by: Claude Sonnet 5.5 <[email protected]>

  • Ben Schumacher(10 Nov 26)

    Fix Support Packet download in Commercial Support modal and move it to Client4 (#39739) * Fix download error handling and support link in Commercial Support modal - Handle network failures and non-JSON error responses so the download button no longer spins forever; show a generic error instead. - Build the error text from message and detailed_error without "undefined" or a trailing ": ". - Revoke the object URL after the download starts so the zip isn't kept in memory until reload. - URL-encode the user's name and email in the cloud support link. - Abort an in-flight download when the modal unmounts. - Drop the incorrect Content-Type header on the GET request. - Rename genereateDownloadURLWithParams to generateDownloadURLWithParams. - Add a key to the support link in the intro text to fix a React key warning that failed the snapshot test. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx2g7qdNkzrw6N9BWos * Remove Commercial Support modal snapshot test The full-modal snapshot only covered the default render and broke on any markup change; the behavior tests cover what matters. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx2g7qdNkzrw6N9BWos * Test the logging banner in the Commercial Support modal Cover when the banner shows (file logging off, or level not DEBUG), when it doesn't (DEBUG, Cloud), its link to the logging settings, and dismiss. Add keys to the banner's <strong> and link elements; without them React logs a "unique key" warning, which fails the test. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx2g7qdNkzrw6N9BWos * Keep %20 spaces and drop stray spaces in cloud support link - Encode spaces as %20 as before, so pages decoding with decodeURIComponent don't show "+" in the name. - Build the name from non-empty parts only. - Test the non-Cloud support link and an empty last name. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx2g7qdNkzrw6N9BWos * Download the support packet through Client4 Add Client4.getSupportPacket, modelled on generatePostExposureReport, and use it in the Commercial Support modal instead of a raw fetch. The request now carries the same auth headers as other API calls (fixing bearer-token clients), and Client4's existing error handling replaces the modal's own JSON parsing and error formatting. Cleanups: - Build the cloud support link with buildQueryString instead of URLSearchParams plus a "+" to "%20" fix-up. - Use a static key for the banner's <strong> tags; react-intl re-keys chunks by index, so the counter wasn't needed. - Inline the generic error message like the file's other strings. - Tests mock Client4.getSupportPacket, share a render-and-download helper, and use test.each for error and banner cases. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx2g7qdNkzrw6N9BWos * Cancel the support packet download when the modal closes Closing the modal hides it first and only unmounts after the close transition, so a download finishing in between still saved the file. Abort in doHide as well as on unmount. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx2g7qdNkzrw6N9BWos * Keep the support packet download going after the modal closes, and read zip bodies regardless of Content-Type parameters Closing the Commercial Support modal no longer cancels an in-flight Support Packet download, matching the behaviour before the move to Client4. The file is still saved; once the modal has unmounted, only the state updates are skipped. doFetchWithResponse now reads the body as a blob when the media type is application/zip or application/octet-stream, ignoring case and parameters such as "; charset=binary". Previously any Content-Type other than exactly "application/zip" was decoded as text, which corrupted the zip. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FpCfx5GWMntpSoAVa1yGrh --------- Co-authored-by: Claude <[email protected]>

Mattermost Security

7.5/10

Repo Security Summary

Updated 24 Aug 26

  • Code-Review10/10
  • CII-Best-Practices0/10
  • Maintained10/10
  • Dangerous-Workflow10/10
  • Security-Policy10/10
  • License9/10
  • Signed-Releases10/10
  • Branch-Protection4/10
  • Token-Permissions0/10
  • Binary-Artifacts10/10
  • Packaging10/10
  • Pinned-Dependencies7/10
  • SAST10/10
  • Fuzzing0/10

Mattermost Website

Website

Mattermost | Operational and AI Sovereignty for Critical Infrastructure and National Security

Zero-trust collaboration, automation and AI in private cloud, self-hosted, and air-gapped environments. Built for national security, critical infrastructure, and regulated enterprises.

Redirects

Does not redirect

Security Checks

All 65 security checks passed

Server Details

  • IP Address104.18.1.109
  • LocationSan Francisco,California,United States of America,NA
  • ISPCloudFlare Inc.
  • ASNAS13335

Associated Countries

  • USUS

Safety Score

Website marked as safe

100%

Trackers

Loads 2 third-party trackers

  • googletagmanager.comGoogle Ads
  • qualified.comqualified.com

Blacklist Check

mattermost.com was found on 0 blacklists

  • AntiSocial Blacklist
  • Artists Against 419
  • Badbitcoin
  • Bambenek Consulting
  • CERT Polska
  • CoinBlockerLists
  • CRDF
  • CryptoScamDB
  • EtherAddressLookup
  • EtherScamDB
  • Fake Website Buster
  • MetaMask EthPhishing
  • NABP Not Recommended Sites
  • OpenPhish
  • PetScams
  • PhishFeed
  • PhishFort
  • Phishing.Database
  • PhishStats
  • PhishTank
  • Phishunt
  • RPiList Not Serious
  • Scam.Directory
  • SecureReload Phishing List
  • Spam404
  • StopGunScams
  • Suspicious Hosting IP
  • ThreatFox
  • ThreatLog
  • TweetFeed
  • URLhaus
  • ViriBack C2 Tracker

Website Preview

Website preview

Mattermost Android App

APK Info

  • AppMattermost
  • Creation Date15 Sept 18
  • Last Updated19 Jul 24
  • Current Version1.11.0
  • CreatorMattermost
  • Downloads50,000+ downloads
  • Privacy ReportView on Exodus →

De-Googled Compatibility

Native3.00/ 41 ratings
microG3.00/ 41 ratings

Tested on Android 15 · Updated 11 May 25 · View on Plexus →

Trackers

  • Sentry

Permissions

  • Access Network State
  • Camera
  • Internet
  • Read App Badge
  • Read External Storage
  • Receive Boot Completed
  • System Alert Window
  • Use Fingerprint
  • Vibrate
  • Wake Lock
  • Write External Storage
  • Update Count
  • Receive
  • Send
  • Read Settings
  • Update Shortcut
  • Change Badge
  • Read Settings
  • Write Settings
  • Update Badge
  • C2d Message
  • Read Settings
  • Write Settings
  • Read
  • Write
  • Broadcast Badge
  • Provider Insert Badge
  • Badge Count Read
  • Badge Count Write

Mattermost iOS App

App Info

Mattermost

Requires Mattermost Server v11.7.0+. Older servers may not be able to connect or have unexpected behavior. ------- Mattermost is the leading collaborative workflow platform for defense, intelligence, security, and critical infrastructure. The platform runs on-premises and in private clouds, delivering secure messaging, file sharing, workflow automation, audio/screenshare, and project management—all with full data and operational control. - Securely share messages, files, and links across from your preferred mobile devices. - Start 1:1 and group voice calls with a single click without leaving Mattermost. - Keep critical workflows moving forward with Playbook. - Get the message out without having to manage every endpoint with Enterprise Mobility Management (AppConfig) support. - Private mobility with ID-only push notifications to keep sensitive data protected while keeping stakeholders informed. To use this app, you need a URL for a Mattermost server. ------- Deploy your own server: https://docs.mattermost.com/guides/deployment.html Terms of Service: https://mattermost.com/terms-of-service/ Contribute to the project: https://github.com/mattermost/mattermost-mobile

Rating

Rated 4.76 out of 5 stars by 5,646 users

Version Info

  • Current Version2.44.0
  • Last Updated16 Sept 26
  • First Released10 Jul 17
  • Minimum iOS Version16.0
  • Device Models Supported130

App Details

  • IPA Size93.41 Mb
  • PriceFree (USD)
  • Age Advisory4+
  • Supported Languages21
  • DeveloperMattermost, Inc.
  • Bundle IDcom.mattermost.rn

Screenshots

  • App screenshot
  • App screenshot
  • App screenshot
  • App screenshot

Mattermost Docker

Container Info

mattermost

Mattermost - open-source team chat: channels, threads, DMs, voice calls, slash commands, integrations, and plugins. Team Edition image on a managed Postgres - it listens on :5000 directly, so no proxy.

#Business Apps

Run Command

docker run -d \
  -e MM_SITE_URL=${MM_SITE_URL} \
  

Compose File

version: 3.8
services:
  mattermost:
    environment:
      MM_SITE_URL: "http://localhost:5000"

Environment Variables

  • Var NameDefault
  • MM_SITE_URLhttp://localhost:5000

Mattermost Reviews

More Team Collaboration

  • Privacy-focused messenger using the Matrix protocol. The Element client allows for group chat rooms, media sharing, voice and video group calls. End-to-end encryption by default for message content, but not metadata, which remains visible to the home server operator.

    Security Audited Open Source element-hq/element-web
  • Easy-to-deploy, self-hosted Slack alternative with polished cross-platform apps. End-to-end encryption is supported, but off by default - without it, workspace admins can read all messages. Usage statistics are sent by default.

  • A decentralized, encrypted P2P friend-to-friend app for chat, forums, channels, boards, voice calls and anonymous file sharing. Best for trusted contacts, as it isn't anonymous unless you route it through Tor (off by default).

  • Open source team chat organized around topic-based threading, which keeps busy conversations easier to follow than channel-only tools. Can be self-hosted, or used as a paid cloud service. The threading model takes some getting used to.

About the Data: Mattermost

Change History

  • Amended (description, openSource, url) by @lissy93 #770
  • Amended (androidApp, iosApp)

Edit Mattermost Data

You can edit Mattermost's entry in this section of awesome-privacy.yml by submitting a PR to our GitHub repo.
Note that some of the information shown above has been aggregated from external sources, a list of these can be found data documentation.

Origin Data

Modify Data

API

You can access Mattermost's data programmatically via our API. Simply make a GET request to:

https://api.awesome-privacy.xyz/v1/services/mattermost

The REST API is free, no-auth and CORS-enabled. To learn more, view the API Docs or read the API Usage Guide.

Share Mattermost

Help your friends compare Team Collaboration, and pick privacy-respecting software and services.
Share Mattermost and Awesome Privacy with your network!