A fresh Activepieces install is generic: default branding, one project, SMTP half-configured, everything running in a single container. These ten customizations make it yours — setup, data modeling, branding, and ops hardening. Every one is a request you can type to the vibe-code agent while it reads your actual server’s files and configuration.

1. Make email actually work

Half of every automation is “email someone,” and it all fails quietly until SMTP is configured. The agent edits the environment file with your real SMTP host, port, credentials, and sender address, restarts the stack, and proves it with a test message. Type: “Configure Activepieces to send mail through our mail server with these credentials, set the from address to automation@company.com, and send me a test email.” Until this works, password resets and flow notifications silently vanish. Do it first, before any of the flows below.

2. Scale out with workers and queue mode

The default single container runs flows in-process — fine for a dozen runs a day, shaky for thousands. The agent can split your compose file into an API container plus dedicated worker containers backed by Redis, so a heavy batch of flow runs doesn’t starve the web UI. Type: “Split my install into separate API and worker containers with Redis, so a big batch of flow runs doesn’t slow down the interface.” On a small server this may be unnecessary. Ask the agent to check your actual run volume first — a good one will say so honestly instead of building infrastructure you don’t need.

3. Put your brand on it

It’s your server; the login page shouldn’t greet your customers with someone else’s logo. Set the frontend URL to your own domain, update the instance name, and adjust the appearance settings so invites to stakeholders look like your company, not a vendor. Type: “Point the frontend at automations.company.com, change the instance name to ‘Acme Automation’, and swap in our logo.” Honest note: deep white-labeling capabilities vary by edition and version — the agent will check what your install actually supports and tell you, instead of promising a rebrand your build can’t do.

4. Organize projects and folders the way teams think

One flat project becomes a junk drawer within months. Projects map to teams or clients, folders map to jobs, and each project can carry its own connections. The agent reads your existing flows and proposes a structure, then moves things and renames them sensibly. Type: “Create a project for Marketing and one for Operations, move each existing flow to where it belongs, and name folders by job.” Permissions follow projects, so this is also an access-control decision — think for a minute about who should be able to open the billing flows.

5. Model flow data in Tables, not spreadsheets

Activepieces has Tables — real databases inside the app that flows can read and write with typed fields. If flows use a shared Google Sheet as their scratch database, they’re one accidental edit away from corrupted data and one permissions change away from a wall of failed runs. The agent designs tables with proper field types (text, number, date, dropdown) and rewires your flows to use them. Type: “Create a ‘Customers’ table with name, email, plan, and renewal date fields, and switch the onboarding flow to write there instead of the ‘Customers’ sheet.” Flow-owned data belongs in the flow platform; human-shared data can stay in Sheets.

6. Lock down who gets in

An automation server is a skeleton key to your business apps — anyone with a login can build a flow that reads your mail or writes to your CRM. Turn off open registration, review who already has access, and separate admin from member roles: admins manage connections and settings, members build flows. Type: “Disable public sign-up on my Activepieces, list every current user with their role, and demote anyone who shouldn’t be an admin.” If your edition supports SSO, the agent can wire that too. Do this before you invite the wider team, not after the first incident.

7. Audit enabled pieces and stored connections

Your install can have hundreds of pieces enabled, each a door to an external API, plus connections holding live OAuth tokens. The agent lists which pieces are actually in use, which connections haven’t run in months, and can disable or prune the dead weight. Type: “List all enabled pieces and all stored connections, show me the last time each connection was used, and disable anything untouched for 90 days.” While it’s in there, have it scan Code pieces for hardcoded API keys — those belong in connections or environment variables, where you can rotate them without editing flows.

8. Housekeeping: backups and log retention

Run history grows without bound in the Postgres database, and the box that backs up the rest of your business deserves its own backup. The agent sets up a nightly job: dump the database, export your flows as JSON into a folder your server backup already covers, and prune run logs past a retention window. Type: “Back up the Activepieces database nightly, export all my flows as JSON to the backup folder, and delete run logs older than 90 days.” Then test the restore once. An untested backup is a rumor, and flow JSON exports are what make a rebuild take hours instead of weeks.

9. Build a custom piece for your internal API

Your internal CRM or ERP has an HTTP API that no built-in piece covers, so flows end up with raw HTTP steps full of URLs and headers. The agent can scaffold a proper custom piece with the TypeScript pieces SDK: named actions, friendly form fields, your auth scheme — then register it in your install so it appears in the piece picker like any built-in. Type: “Build a custom piece for our internal orders API with actions for ‘create order’ and ‘get order status’ using our API-key auth, and install it on my server.” Honest effort note: this is more work than a Code step — expect an hour of iteration and testing — but your business users get a clean labeled action instead of a wall of JSON.

10. Watch the server’s health like it’s production

Treat the automation box as production, because it is: container health checks, restart policies that recover from a crash, disk-space alerts (logs and run data only grow), and telemetry off if you prefer not to phone home. The agent wires health checks and always-restart policies into your compose file and adds a scheduled disk and health check with an alert channel. Type: “Add health checks and always-restart policies to my Activepieces containers, alert me if disk usage passes 80%, and turn off telemetry.” The worst failure mode of an automation server is a dead one that nobody noticed for a week. Make that loud.

Unglamorous first

Do the boring ones — SMTP, access control, backups — before the fun ones; everything else stands on them. And if you’d rather receive the whole thing already configured and hardened, that’s exactly what the product offers: Activepieces on OpenSysLab.