One-off jobs are nice; the real payoff is workflows that run on their own. The agent can write scripts, schedule them on your server, query the Docmost database, and wire it to your other tools. Here are ten multi-step workflows worth automating — each ends with less manual checking, not more.

1. Nightly backups you have actually restored

A backup you have never restored is a hope, not a backup. The agent writes a script that dumps the Docmost Postgres database and archives the uploads folder into a timestamped bundle, keeps 30 days, and syncs a copy off-site if you have S3. You get a short email only when a run fails; silence means it worked. Then it schedules a monthly restore into a scratch database to prove the dump loads. The first real restore almost always exposes a gap — better at 9am on the first of the month than during an incident. Try: “Back up the Docmost database and uploads every night at 2am, keep 30 days, and restore one into a test database on the first of each month.”

2. A weekly “what changed” digest

In a busy wiki, nobody reads every edit. The agent queries the page-history tables for the past seven days and emails you one digest per space: page title, author, and a one-line summary of what moved. Monday stops starting with twenty minutes of clicking through the sidebar, and you notice the quiet edit that rewrote a policy three weeks ago. Try: “Every Monday at 8am, email me everything edited last week in the Engineering and Policies spaces, grouped by author.”

3. Stale-page patrol

Documentation rots quietly. The agent schedules a monthly sweep for pages with no edits in 120 days, then posts a comment on each one naming the owner and asking them to confirm or update it — a nudge that appears exactly where people already work. Pages nobody will maintain get flagged so you can delete them honestly instead of leaving landmines for new hires. Try: “Every month, find pages nobody has touched in four months and comment on each asking the owner to review it.”

4. Comment notifications in Slack

Docmost comments are where review happens, but people have to open the wiki to see them. Because the agent can read the comments table, it polls on a schedule and posts new comments to a Slack channel through a webhook — page title, author, a short snippet, and the link. No plugin exists for this; a cron job and a webhook is the whole integration. Polling every few minutes is plenty for review comments, and the agent can throttle noisy spaces so one active thread does not drown the channel. Try: “When someone comments on any page in the Support space, post it to our #docs Slack channel with a link.”

5. A draft-to-public publish pipeline

Keeping public docs in sync with the internal wiki is usually copy-paste drudgery. Mark pages ready with a “publish” label, and the agent’s scheduled job exports them to HTML and uploads them to your public docs folder — or flips on public sharing if you publish through Docmost itself. The internal wiki stays messy; the external one stays current. Try: “Every Friday, export everything labeled ‘publish’ to HTML and upload it to our public docs site.”

6. An inbox folder that becomes wiki pages

Give the agent a watched folder on the server. Drop a Markdown file in, and a scheduled job converts it, creates a page in the right space with a clean title, applies the labels you configured, and links it from the space index page. Sales one-pagers and old spec documents stop living as attachments in chat threads. Try: “Watch /srv/drop-docs — anything I put there becomes a page under Sales Playbooks, labeled and linked from the index.”

7. A Git mirror of your runbooks

Some teams want runbooks reviewed like code. The agent exports the Engineering space to Markdown on a schedule and commits it to a Git repository, so diffs, pull requests, and blame all work on your documentation. Round-tripping edits back into Docmost needs discipline, but as a read-only mirror it is free oversight on every change. If a page vanishes from the export, the diff shows it was deleted in Docmost — usually exactly the review you wanted. Try: “Every night, export the Engineering runbooks to Markdown and commit them to our ops repo.”

8. New-hire onboarding, end to end

Provisioning a wiki account is five clicks everyone forgets. Give the agent the new person’s name and email and it creates the account, adds them to the right groups, builds a Start-here page linking the ten pages they need first, and emails them the login link. The packet comes out identical every time, which is the point. Try: “When I tell you a new support rep starts Monday, create their account, add them to Support and All-Hands groups, and build them a Start-here page.”

9. A compliance archive that exports itself

Auditors ask for documentation as it existed on a specific date. The agent schedules a monthly export of your Policies space to an HTML bundle with timestamps, stored in a dated folder on the server. When the question finally comes, you hand over a folder instead of losing a weekend to archaeology. Each bundle also lists pages created and deleted that month, so the archive tells a story rather than showing a snapshot. Try: “On the first of each month, export the whole Policies space as a timestamped HTML bundle into /srv/compliance-archive.”

10. Attachment and disk housekeeping

Uploads grow until disk space becomes the incident. A weekly job reports the uploads folder size, lists the twenty largest attachments, and flags files attached to pages that no longer exist. Growth trends reach your inbox while they are still boring, which is exactly when you want to hear about them. Try: “Every week, report the uploads folder size, list the 20 largest attachments, and flag any attached to deleted pages.”

Pick the one workflow where manual checking is currently somebody’s job. The agent builds it, shows you a test run, and leaves the script on your server where you can read it. See Docmost on OpenSysLab.