Daily jobs and scheduled workflows are only two of the three layers. The third is the workspace itself: how it looks, how your data is modeled, and how hard it is to break into. These ten customizations are configuration-level work where the agent’s access to your files and settings pays off.

1. Make it look like your company

A default install looks like nobody owns it. The agent replaces the logo and favicon, sets your brand colors, changes the login page background, and writes custom CSS for the details the settings page does not cover. It keeps a copy of the original values so you can roll back, and it works through the Layout settings where these options actually live. None of it touches the application code, so upgrades keep working afterward. Try: “Replace the logo with our brand file, set the primary color to our brand blue, and restyle the login page with the company background.”

2. Model your people and rooms with custom fields

Default profiles carry a name and an email; your directory can carry more. The agent adds custom user fields, department, office, phone extension, and custom fields on omnichannel rooms such as account tier or product area, then verifies they show up on profiles and in search. This is real data modeling, and getting the fields right early saves you from spreadsheet sidecars later. Try: “Add custom user fields for department, office, and phone extension, show them on every profile, and make sure I can find people by department in the directory.”

3. Cut real roles out of the admin-or-user binary

Two roles do not fit a real organization. The agent creates purpose-built roles, a Contractor role that can read and post but cannot create channels or invite people, a Team Lead role that can pin and moderate, an Auditor read-only setup, by working through the granular permission matrix one entry at a time. It shows you the matrix before applying, so the new role does what you think it does. Try: “Create a Contractor role that can read and post but cannot create channels, invite users, or start direct messages.”

4. Set message retention and audit rules

Keeping everything forever is a liability, and deleting everything on a whim is another. The agent configures the retention policy with global defaults and per-channel overrides, prunes old file uploads, and restricts who can delete messages. Retention then runs on a schedule inside Rocket.Chat rather than as a manual purge someone keeps forgetting. If your edition includes the audit panel, it enables that too, so you can answer who said what and when without digging through the database. Try: “Keep public channel messages for 12 months and direct messages for 90 days, delete files older than 6 months, and make message deletion a moderator-only power.”

5. Turn on E2EE where it matters, eyes open

End-to-end encryption protects a room’s contents even from the server admin, which is exactly why it has trade-offs. Search stops working inside encrypted rooms, bots and integrations go blind, and a lost key means lost history. The agent enables E2EE for the rooms you name, sets expectations with the team, and explains plainly what stopped working before you roll it out wider. Try: “Enable E2EE for the #leadership channel, walk me through what stops working in encrypted rooms, and draft the note we send the team.”

6. Build omnichannel like a real support desk

Departments, agents, managers, canned responses, business hours, and SLA policies are the skeleton of customer messaging, and most installs never set them up. The agent builds the structure: departments for sales and support, the right agents and managers in each, business hours, and canned answers for the ten questions your team types daily. A few of these features are edition-gated, and the agent will tell you which ones your version has. Try: “Set up two omnichannel departments, sales and support, put the right agents in each, set business hours of 9 to 6, and create canned responses for pricing, refunds, and shipping.”

7. Harden the front door

An internet-facing chat server gets probed every day. The agent locks the obvious doors first: two-factor required for admins and moderators, a real password policy, rate limiting on the REST API, registration closed to invites only, and, if you have fixed office IPs, admin settings reachable only from your network. Each change is one setting it can show you afterward. It also checks the small things owners miss, like whether open registration is really off and whether any admin account still runs on a weak password. Try: “Require 2FA for admin and moderator accounts, set a strong password policy, rate-limit the API, and make registration invite-only.”

8. Wire in video calls the way you want them

Rocket.Chat’s call button can start a Jitsi or BigBlueButton room instead of leaving everyone hunting for a link. The agent installs the integration, points it at a free public Jitsi server to start, or at your own self-hosted instance if calls should stay private, and sets the default for channels and DMs. One click from a chat message turns into a working call. Try: “Connect our self-hosted Jitsi server as the default call provider for channels and direct messages.”

9. Open the federation door when partners need in

If customers or suppliers run their own chat servers, Matrix federation lets a room span both sides without guest accounts or email chains. The agent does the fiddly part, domain validation, the DNS records federation checks, and the workspace settings, then enables it for the channels you choose and tests with an external account before you announce it. Try: “Set up federation on chat.ourcompany.com, get the DNS validation passing, enable it for #partners, and test it against an external Matrix account.”

10. Prove your backups with a restore test

A backup that has never been restored is a rumor, not a plan. The agent schedules a nightly mongodump of the Rocket.Chat database plus the uploads folder, keeps a sensible number of old dumps, and then actually restores one into a scratch database to prove the files are good. It leaves you a short runbook so a restore does not depend on one person’s memory. The restore test deserves its own scheduled slot, because a backup process you never exercise is exactly the one that fails the day you need it. Try: “Set up nightly MongoDB backups that keep 14 days, include uploaded files, and run a restore test this week to prove it works.”

See it in action

Official walkthroughs from the Rocket.Chat team:

Worth a watch next:

• Migrate from Rocket.Chat Cloud-hosted to Self-manged Workspace — moving a workspace onto your own server, the deepest customization there is.
• Managing Your Cloud Hosted Rocket Chat Workspace — an admin tour of the workspace settings you’ll extend with custom fields and roles.

These changes are deeper than daily jobs, so take them one at a time and let the team adjust between them. The agent does the work on your server and shows its steps, and you keep the final say. See Rocket.Chat on OpenSysLab.