Activepieces: how it works, credits, cloud vs. self-hosted, and limits

Activepieces in three lines

Activepieces is a no-code automation tool: “flows” connect your apps, with AI agents, tables and an MCP server around them. The vendor, Activepieces Inc., is a US company; the core of the software is MIT-licensed. It comes as a cloud service hosted in the European Union and as a version you install yourself. It bills credits: one per flow run.

This page analyzes its public pricing and documentation as recorded on the date shown. Nobody here has run the product.

What Activepieces actually does

A flow starts from a trigger and chains steps. Connectors are called “pieces”. On how many there are, the vendor says: 775 "pieces" and 6,759 actions according to the pieces page; "770+" on the pricing page source. That is fewer than the catalogs announced by Zapier, Latenode or Pabbly Connect, so check that your apps are on the list before anything else.

Around flows, the free plan already includes agents, tables (data storage), a chat and API access. The number of flows isn’t capped: Unlimited flows on every plan.

On the cloud, the documentation sets a few technical limits: up to 10 minutes per run (waiting and approval pauses don’t count), run data kept for 30 days, and a number of concurrent runs per project that depends on the plan (5 on Free, 10 on Plus, 20 on Team). Past that, runs wait in a queue.

What Activepieces bills: one credit per run

The rule: one flow run = 1 credit, however many steps it has. Take a 5-step automation that runs 1,000 times a month:

• Activepieces: 1,000 credits, exactly the free plan’s quota (1,000 credits per month);
• the same with 10 steps: still 1,000 credits;
• on Zapier, which counts each successful action: 4,000 tasks for 5 steps, 9,000 for 10;
• on Make, which also counts the trigger: 5,000 credits, then 10,000.

The longer your flows, the more this rule works in your favor. The guide to billing units compares these ways of counting.

Two contradictions to know about:

• The AI scale. The vendor’s pages contradict each other: the pricing FAQ gives the fixed scale above, while the “Breaking Changes” page of the documentation says managed AI calls are now billed at the provider’s actual cost, converted into credits, and that the fixed cost per model is gone.
• How free credits renew. The Free plan card announces 1,000 credits a month; the table on the same page says the reset is daily. Our database keeps 1,000 credits per month.

Once the quota is used up, what happens depends on the plan. On Free, flows stay switched on but extra runs don’t go through until the credits renew. On Plus and Team, flows keep running and each extra credit is billed (the rate is in the box above).

The plans at a glance

Table scrolls horizontally →

Activepieces plans

Plan	Monthly price	Monthly price, billed annually	Monthly quota included	Limits
Freesource	$0	$0	1,000credits	Up to 5 users, unlimited flows, agents, chat, tables, MCP, API and SDK; community support. Beyond the quota, runs wait for the credits to renew.
Plussource	$20	$16	10,000credits	Flat price per month, up to 5 users, 10,000 credits per month reset every month, then per-credit billing. Personal AI keys, community support. 10 concurrent runs per project on Cloud.
Teamsource	$200	$166	50,000credits	Flat price, 25 users included, 50,000 credits per month, then per-credit billing. Unlimited projects, SSO and standard roles, global connections, email support, payment by ACH or wire transfer possible. 20 concurrent runs per project on Cloud.
Ultimatesource	custom pricing	custom pricing	—	Custom pricing ("Talk to sales"). Credits sized to order as an annual allowance, unlimited users, dedicated workers, execution priority, SCIM and custom roles, Releases and Git Sync, audit logs, secret managers. Custom branding, custom billing, dedicated support.

• Free: 1,000 credits per month, up to 5 users, no credit card.
• Plus: $20 per month billed monthly, $16 per month billed annually, for 10,000 credits. Flat price for up to 5 users, personal AI keys.
• Team: $200 per month for 50,000 credits, 25 users included, unlimited projects, SSO, email support.
• Ultimate: custom pricing, with an annual credit allowance.

Prices are flat, not multiplied by the number of users within the stated limit. The pricing page advertises two months free on annual billing and mentions no free trial of the paid cloud plans. Two plans outside the grid exist for white-label embedding and large enterprises; they are listed in the notes at the bottom of this page.

The cost calculator turns your volume into credits and compares it with other tools.

Cloud or self-hosted

On self-hosting, the vendor’s answer is: Yes: free Community Edition with no credits; every plan on the pricing page is also available self-hosted source. On the license: MIT ("MIT Expat") for the core; commercial license for the Enterprise edition folders.

Self-hosting doesn’t mean “no credits” in every case. The documentation separates two editions:

• the Community Edition, free and with no credits, which the pricing FAQ describes as the automation core on its own; it no longer supports multiple projects;
• the Enterprise edition, which the Docker Compose script and the Helm chart install by default. Without a key, it runs on the free plan, with its credit limits; Plus and Team are subscribed to from inside the app.

The vendor’s pages contradict each other on one point: the Free plan card lists agents, while the FAQ places running agents and chat among the things paying adds to the Community Edition.

On hardware, the documentation asks for Docker Compose v2 and at least 2 vCPU and 4 GB of RAM, with PostgreSQL and Redis in the recommended stack. When you self-host, the cloud’s technical limits are set with environment variables. The guide Activepieces: cloud or self-hosted covers the choice between the two.

Where your data lives

For the cloud: Cloud hosted in the EU, in Frankfurt by default; another region on request; self-hosting in the region of your choice source. If you have European customers and want their data kept in the EU without running a server, that is what separates Activepieces from Zapier, which stores data in the United States.

AI, agents and MCP

The AI features on record: Agents, Chat and AI Router; AI steps at 2, 10 or 20 credits depending on the model, 1 credit with your own keys. MCP is a protocol that lets an AI assistant act in your tools (see the MCP guide); here: Built-in MCP server; "Free on every plan, cloud or self-hosted" according to the vendor.

Alternatives worth a look

• The other tool you can install yourself, also billed per execution: n8n, and n8n alternatives.
• If the app catalog comes first: Zapier, and Zapier alternatives.
• The wider picture: workflow automation tools compared.