Activepieces

What it is

Activepieces is a workflow automation platform you can run on your own infrastructure under the MIT licence. The core is genuinely MIT at github.com/activepieces/activepieces; the enterprise features live in packages/ee and packages/server/api/src/app/ee under a separate commercial licence, and running those on a self-hosted instance takes a paid key. That split matters more than it sounds, and the Versus section below is where it bites.

The catalogue is the reason most teams arrive. Activepieces ships 760+ integrations — it calls them pieces — covering Slack, Gmail, Salesforce, HubSpot, Notion and the model providers. Every piece is exposed as an MCP server automatically, so the same connector you use inside a flow is callable from Claude Code, Cursor, Windsurf or Zed without writing a server. Activepieces is also an MCP client, which means a flow can call out to external MCP servers. Both directions, one deployment.

Install is curl -fsSL https://get.activepieces.com | sh on Docker Compose with PostgreSQL and Redis. A Helm chart exists for teams already running Kubernetes.

The fork it resolves

The question this page answers is what you run after Zapier’s per-task meter stops making sense, or after you read n8n’s licence properly.

Zapier meters per task, and a task is a step. A 20-step Zap that fires 500 times a month burns 10,000 tasks, and premium apps like Salesforce and HubSpot double the consumption. Activepieces bills 1 credit per run no matter how many steps are in it. A two-step flow and a twenty-step flow cost the same. That single design decision is the whole pitch to anyone whose Zapier bill grew because their flows got better, not because their volume did.

The n8n fork is different and less well understood. n8n is not open source and does not claim to be — it ships under the Sustainable Use License, a source-available fair-code licence that restricts use to internal business, personal or non-commercial purposes. You cannot charge people for access to an n8n instance you run, and you cannot sell a white-labelled version. If your plan is to embed automation inside a product you sell, n8n’s licence is the wall you hit and Activepieces’ MIT core is the way around it. Activepieces sells embedding as a first-class enterprise feature: the builder inside your app, under your brand.

Activepieces is also the open-source landing spot for Pipedream refugees. Workday signed the agreement to acquire Pipedream on 19 November 2025, expected to close by 31 January 2026, and folded it into a Workday-owned agent platform alongside Sana and Flowise. If your automation layer is now a line item inside an HR suite’s roadmap, MIT plus self-hosting is the structural answer.

Pricing reality

Free is $0 with 100 credits per day, single user, self-hosting at the same price. Plus is $16 a month billed annually with 10,000 credits a month and up to 5 users. Team is $166 a month billed annually with 50,000 credits and 25 users, and it is where SSO and standard roles turn on. Ultimate is custom, and it carries SCIM, custom roles, audit logs, secret managers, event streaming and dedicated workers. Overage on Plus and Team is $0.007 per credit.

Read the credit table before you model the bill, because the plumbing is nearly free and the AI is the entire cost. One credit covers a whole run however many steps it has. On top of that, an agentic action costs 1 credit, a fast model step 2, a smart model step 10 and a frontier model step 20. Connect your own AI provider key on Plus or above and every AI step drops back to 1 credit, with the model bill going to the provider directly.

Work that through. On Plus, 10,000 credits is 10,000 non-AI runs a month at $0.0016 each. Put one frontier-model step in the same flow and that allowance buys 500 runs. Bring your own key and it buys 5,000. The gap between the worst and best configuration of the same flow is 10x, and it is a checkbox.

Self-hosting is $0 for the MIT core. Our estimate from the documented Docker Compose requirements — one small VM, managed PostgreSQL, Redis — is $40-80 a month on a mainstream cloud, before anyone’s time.

Best for

The ops or platform engineer at a 20-100 person company that has outgrown Zapier’s task meter and needs the automation layer inside its own perimeter. The scoped case where Activepieces wins outright: a RevOps team running enrichment and routing flows with 15-25 steps each, several thousand runs a month, using its own Anthropic or OpenAI key, with the whole thing self-hosted because legal will not sign off on customer records transiting a vendor’s queue.

It is also the correct pick when you are embedding automation into a product you sell. MIT plus a documented embed path is a combination n8n’s licence does not offer.

Not for

A five-person team with no one who owns Docker. Cloud Free is a single seat, so the moment a second person needs to edit a flow you are on Plus or self-hosting, and self-hosting means somebody owns PostgreSQL backups and upgrades.

It is also the wrong pick if your buying committee weights vendor scale. Activepieces is a Y Combinator-backed company founded in 2022 with a headcount that aggregator trackers put somewhere between 11 and 23. n8n raised a $180M Series C in October 2025 at a $2.5B valuation. If procurement requires a vendor that will outlive the contract on balance-sheet evidence, this is a hard sell — see the first watch-out for the guard.

Versus the alternatives

• Zapier — the largest installed base in the category and the fastest path to a working automation, from $19.99/mo. Pick it when the flows are short, the volume is low, and nobody on the team wants to think about infrastructure. Its per-task meter is the reason people leave, not the reason they arrive.
• Make — the other high-share incumbent, from $9/mo, with a visual builder that handles branching and iteration better than Zapier’s. Pick it when the work is data transformation with heavy branching and cloud-only is acceptable. It has no self-host path at all.
• n8n — the fastest-growing entrant in the segment: $180M Series C at a $2.5B valuation in October 2025, ARR past $40M, 1,400+ enterprise customers. Pick it when you want the deepest AI-agent node set and the largest template library, and when your use is internal so the Sustainable Use License costs you nothing. Its community and hiring pool are both larger than Activepieces’.

If none fit, write the integration as code against the four APIs you actually touch and run it on the scheduler you already operate. Under roughly ten flows, a platform is overhead.

Watch-outs

• The company is small and the licence is your insurance policy. Aggregator headcount estimates sit between 11 and 23 against n8n’s funded scale, and revenue estimates in the low single-digit millions. Guard: self-host from day one rather than starting on cloud, keep a scheduled pg_dump of the PostgreSQL database off-instance, and pin an image digest. The MIT core means an acquisition cannot take the software away from you — but only if you already hold a running copy and a restore you have rehearsed.
• Enterprise features are commercially licensed even when you self-host. SSO lands at Team, SCIM, audit logs and custom roles at Ultimate, and the packages/ee code is not MIT. Guard: confirm which of SSO, SCIM and audit logs your security review mandates before the pilot, and get the Ultimate quote then. Discovering the requirement after the flows are in production means negotiating from the weaker side.
• The AI credit multiplier is a 10x swing on a checkbox. A frontier model step costs 20 credits; the identical step on your own key costs 1. Guard: connect your own Anthropic or OpenAI key before you build the first AI flow, and set a credit alert at 70% of the monthly allowance so the overage at $0.007 per credit shows up as a warning rather than an invoice.
• 760+ pieces is a community catalogue, and depth varies. The count includes contributed pieces, and a piece can cover the three endpoints its author needed rather than the API surface you need. Guard: before committing a flow to production, open the piece in the repo and check that the specific actions and triggers you depend on exist — the pieces are in-tree, so this is a five-minute read, not a support ticket.