Self-Host Activepieces as Your Zapier Alternative

By the end of this playbook you will have:

• Activepieces self-hosted with PostgreSQL + Redis (the production path — not the hobbyist single-container embed)
• HTTPS (or at least a stable AP_FRONTEND_URL) so webhooks work
• Secrets (AP_ENCRYPTION_KEY, AP_JWT_SECRET, DB password) backed up offline
• A first Zapier-replacement flow: CRM/form event → Slack alert (or Sheet append)
• Sign-up locked down after the first admin; workers visible in Platform Admin

Official install: Self Host Activepieces (Docker Compose). Prefer that over random blog compose files.

1. VPS — Activepieces asks for roughly 2 vCPU / 4 GB RAM minimum for the Compose stack.
2. Docker Compose v2 (docker compose version must work — Compose v1 will not).
3. Domain (recommended) — e.g. flows.yourdomain.com A-record → VPS. Webhooks need a URL third parties can reach.
4. Ports 80/443 open if you terminate TLS on-box; or terminate at Cloudflare and proxy to host port 8080.
5. Accounts for the first flow: Slack workspace + HubSpot/Pipedrive/Google Sheets (whatever you are replacing in Zapier).
6. Password manager for .env secrets.

SSH to the VPS, then in the directory where you want the stack:

curl -fsSL https://get.activepieces.com | sh

Options from the docs:

1. 1

   Port and Directory Options

   --port 8090 if 8080 is taken
   --dir /opt/activepieces to choose the folder

2. 2

   Four Containers Started

   The script writes config, generates secrets, and starts four containers: app, worker, postgres, redis.

3. 3

   Manual Install Path

   Prefer not to pipe a remote script? Use the manual path in the same docs page: clone activepieces/activepieces, run sh tools/deploy.sh, verify secrets with grep, then docker compose -p activepieces up -d.

docker compose -p activepieces ps
curl http://localhost:8080/api/v1/health

All four should be Up. Open http://SERVER:8080 (or your HTTPS URL).

1. Create the first account — it becomes platform admin (no default password).
2. Open Platform Admin → Operations → Workers. You must see at least one worker.
3. If Workers is empty: worker cannot reach the app. Official fix — give the worker service AP_FRONTEND_URL=http://app while the app’s public AP_FRONTEND_URL stays your real domain (see Activepieces troubleshooting).

1. In .env, set AP_FRONTEND_URL to https://flows.yourdomain.com (no trailing path surprises).
2. Put a reverse proxy (Caddy, Traefik, Nginx, or Cloudflare) in front of host port 8080 / AP_HOST_PORT.
3. Restart: docker compose -p activepieces up -d.
4. Confirm webhooks: create a test Webhook trigger piece and POST to the public URL.

Goal: when a new HubSpot deal (or Google Form → Sheet row) appears, post a Slack message to #revops.

1. 1

   Build First Business Flow

   In Activepieces, Create Flow.
   Trigger: HubSpot “new deal” (or Schedule + Sheet “new row” if CRM OAuth is blocked today).
   Action: Slack “Send channel message” with deal name, amount, owner, link.
   Connect accounts via Activepieces connections UI (OAuth) — never paste long-lived tokens into piece fields when OAuth exists.
   Publish/enable the flow.
   Create a test deal named AP TEST and confirm Slack fires once.

2. 2

   Migrate Next Zaps

   That is your first migrated Zap. Next migrations: lead routing, enrichment, and approval-gated updates (add a human approval piece or route writes through n8n HITL if you already standardized there).

• [ ] Health endpoint OK; four containers Up
• [ ] Workers page shows ≥1 worker
• [ ] AP_FRONTEND_URL is HTTPS public URL
• [ ] .env backed up offline (especially encryption key)
• [ ] First Slack (or Sheet) flow published and tested
• [ ] Open sign-up disabled (or SSO only)
• [ ] Postgres dump job exists
• [ ] Owner documented for upgrades

• Hobbyist PGLITE in production — dead end for teams; start on Compose.
• Losing AP_ENCRYPTION_KEY — connections become permanent garbage.
• Worker AP_FRONTEND_URL=localhost — empty Workers page / socket errors.
• Compose v1 — install fails; use Compose v2.
• ngrok forever — fine for a demo; not a production webhook URL.
• Migrating every Zap day one — move the top 3 revenue-critical Zaps first; keep Zapier as fallback for a week.

Activepieces vs n8n?
Both self-host well. Activepieces leans piece/Zapier-shaped UX; n8n is stronger today for AI Agent + HITL tool gates. Many teams run Activepieces for classic automations and n8n for agentic ops.

Can I keep using Zapier MCP for Claude?
Yes — different layer. Activepieces replaces scheduled/event Zaps; Zapier MCP is chat-time tool calls.

Need a migration plan off Zapier? Solutions · Architecture · Contact.