Huly is built to be reshaped: custom trackers beyond plain issues, custom fields on any type, teamspaces, roles, and a server stack you fully control. Your vibe-code agent does the reshaping — it edits configuration, talks to the databases, and restarts services — so you describe the outcome and check the result. Ten customizations worth asking for, roughly easiest first.

1. Stand up a custom tracker for IT requests

Huly’s real power move is that “issue” is not the only shape work can take — you can define whole new tracker types with their own fields. IT requests are the classic case: hardware, access, and software asks each need different fields than a bug does, and routing them through the Bug tracker corrupts both queues.

Try: “Create a new tracker called IT Requests with three types: Hardware, Access, and Software. Hardware requests need a cost estimate number field and an approval field with Pending, Approved, and Rejected options.” The support queue stops pretending to be engineering work.

2. Add fields that match your triage

Out of the box, Huly gives you status, priority, assignee, and labels — enough for generic work, thin for a real support process. Custom attributes fill the gap: customer impact, affected environment, root cause. The agent can add them through Huly’s class system and make the important ones required, so empty fields cannot hide at report time.

Try: “Add a Customer impact select field to the Bug tracker with options None, Single customer, and Multiple customers, and make it mandatory when severity is set to Blocker.” Reports suddenly have something to group by.

3. Rebuild the label taxonomy

Labels decay. People create ‘bug’ and ‘Bug’ and ‘bugs’, half the labels were used once and abandoned, and nobody remembers what ‘p2-ish’ meant. The agent can audit usage straight from the database, merge synonyms, delete the orphans, and write the agreed list into a wiki page so the taxonomy has a source of truth people can check.

Try: “Audit every label in the workspace with usage counts. Merge obvious duplicates, remove labels used fewer than two times, and write the final list as LABELS.md in the Engineering teamspace.” Boards get readable again.

4. Brand the workspace

Defaults make the workspace feel like a demo. Huly lets you set the workspace name and icon, and depending on your build, the sign-in copy. The agent can apply your name and logo through the settings or configuration files and restart the front end, so the change shows up everywhere at once instead of piecemeal.

Try: “Rename the workspace to Acme Internal and set our logo as the workspace icon; if the login page text is configurable in this build, set the welcome line to ‘Acme engineering HQ’.” A small touch, but a real ownership signal for the team.

5. Set notification defaults people can live with

Every new member starts with default notification settings, and defaults are usually wrong: either silence or a firehose of email until they mute everything. The agent can walk the user records in the database and set a sane baseline — mentions and assignments only — then tell people how to widen it if they want more.

Try: “Set every full member’s notification preferences to email only for mentions and issue assignments, leave admins as they are, and post a note in #general explaining how to change it.” Fewer muted channels, fewer missed pings.

6. Harden the deployment

Self-hosting means the security work is yours. The agent can audit the compose file and environment, move secrets out of plain sight into a properly permissioned env file, lock CORS to your real domain, confirm TLS is terminated correctly, and report anything that looks like a default credential still sitting in place.

Try: “Audit the Huly deployment: list every secret in the compose file, move them into a .env file readable only by root, and make sure CORS and the public URL are locked to https://huly.ourcompany.com.” Ten minutes of review, months of exposure removed.

7. Fix slow search

Global search leans on Elasticsearch, and an unhappy index shows up as spinning wheels in the search box. The agent can check cluster health and shard counts, see whether the index is degraded, rebuild it if needed, and — when the slowness lives in the queries themselves — suggest or add database indexes and measure the difference before and after.

Try: “Searching for ‘onboarding’ in global search takes forever. Check the Elasticsearch cluster and index health, rebuild any degraded index, and time the same search before and after.” Search goes back to being a reflex.

8. Template the setup for every new client

If you run client work, every new engagement means recreating the same skeleton: project, trackers, components, kickoff doc. The agent can capture that skeleton once as a script and stamp it out per client — including seeding the wiki with your standard kickoff pages, so day one no longer starts from a blank workspace.

Try: “Create a template called ‘new client’ with a project containing Bug, Feature Request, and Change Order trackers, components for Frontend, Backend, and Infrastructure, and a kickoff wiki doc — then instantiate it for a client called Northwind.” New engagements start in minutes.

9. Normalize timezones and calendars

A distributed team with wrong timezone settings gets misleading due dates and midnight pings. The agent can set the workspace default, flip the week to start on Monday if that is your world, and audit user profiles for accounts missing a timezone, so the calendar stops lying to people across locations.

Try: “Set the workspace timezone to Europe/Berlin, make Monday the first day of the week everywhere, and list any user account with no timezone set so I can chase them.” Dates start meaning the same thing to everyone.

10. Audit roles and ownership

Over time, permissions creep: ex-contractors still have accounts, too many people hold owner keys, and roles no longer match reality. The agent can pull every account with its role and last activity from the database, flag accounts idle for 90 days, and confirm the owner list matches who should actually hold admin rights.

Try: “List every workspace account with its role and last login from the database. Flag anyone inactive for more than 90 days and make sure only Dana and me are workspace owners.” Cleanup that would take an afternoon becomes a short review.

Customizations are where self-hosting pays: you are not waiting on a vendor’s settings page. Describe the shape you want, let the agent do the plumbing, and keep a backup before the bigger changes. See Huly on OpenSysLab.