Activepieces
Activepieces is an open-source AI automation platform where agents, visual flows, tables and chat share one workspace across 763 app integrations, with enterprise governance and a choice of EU cloud, self-hosted or air-gapped deployment.
What is Activepieces?
Activepieces is an AI automation platform built around an open-source core released under the MIT licence, with cloud and enterprise editions under a separate commercial licence. It is published by Activepieces Inc., a San Francisco company that carries the Y Combinator "Backed company" badge in its footer.
The product puts four building blocks in one workspace. Agents are given a goal in plain language and work out the steps themselves, using whichever connected apps, existing flows, MCP servers or uploaded files you allow them. Flows are the classic automation object: a trigger, then the steps you choose, with branching, loops, retries, waitpoints, sub-flows and human approvals. Tables hold rows and columns that both flows and agents can read and write. Chat, still in beta, builds a flow from a sentence. The four combine freely: an agent can call a flow, a flow can run an agent, and both share the same tables and a searchable Knowledge Base of your documents.
Coverage comes from 763 open-source pieces, the largest groups being artificial intelligence, productivity and marketing, and anything not covered is reachable through the HTTP piece or an inline TypeScript step. The same catalogue is exposed as a hosted MCP server, so an assistant such as Claude, Cursor, Windsurf or Zed can run what you built.
What separates Activepieces from lighter automation tools is the governance layer. Teams get sealed projects; access follows Okta or Entra through SSO and SCIM; roles can be locked down to a single action; every prompt, tool call and payload is recorded and can be streamed to a SIEM; and credentials can live in HashiCorp Vault, AWS Secrets Manager, CyberArk Conjur or 1Password rather than in the product. Model keys are yours as well, whether OpenAI, Anthropic, Gemini, Azure, Bedrock or any OpenAI-compatible endpoint.
Deployment is the third distinguishing point. The same build runs on the Activepieces cloud in Frankfurt, on your own Docker or Kubernetes cluster, or in an air-gapped network with no outbound traffic at all, at the same price.
What it does
- Run an AI agent that takes a job end to end across your connected apps, pausing only where you asked for approval
- Build automated flows from a trigger and a chain of steps, with branching, loops, retries and error handling
- Describe what you want in plain words in Chat and get a working flow built for you
- Connect 763 open-source app integrations, or reach any REST or GraphQL endpoint through the HTTP piece
- Store and share data in Tables that every flow and agent can read and write
- Expose your automations as an MCP server so Claude, Cursor or another AI client can run them
- Deploy the same build on the EU cloud, on your own Docker or Kubernetes infrastructure, or fully air-gapped
When to use Activepieces / When not to
A quick filter to help you decide if Activepieces is the right fit.
When to use Activepieces
- Business teams with no coding background, who describe the job they want done in plain words and let the platform build the flow
- Enterprises that need SSO with Okta or Entra, granular roles, SCIM, audit logs streamed to a SIEM and credentials held in their own vault
- Organisations under data-residency constraints, who can run the product on the Frankfurt cloud, on their own infrastructure, or fully air-gapped
- SaaS vendors who want to embed a white-label automation builder and 763 integrations inside their own product
- Engineering and open-source teams who want to read, fork and extend the code, and write their own integrations in TypeScript
When not to use Activepieces
- Anyone who needs a mobile app: there is no iOS or Android release, and no browser extension either
- Buyers in regulated procurement who need a settled compliance story today, since the security page claims SOC 2 Type 2 while the pricing FAQ says the SOC 2 programme is still in progress
- Teams who want to read the contract before signing up, because the Terms of Service and Privacy Policy pages return "This document is not available right now"
- Very high-volume users on the free plan, which stops at 100 credits a day and makes further runs wait until the next morning
- Self-hosters expecting the full product for free: the Community Edition is the automation core alone, without Agents, Chat, projects, roles or audit logs
How to use Activepieces
A typical end-to-end flow, from setup to results.
- Create a free account on the Activepieces cloud, with no credit card, or self-host the platform with a single Docker command or a Kubernetes deployment
- As an administrator, register your AI provider key once (OpenAI, Anthropic, Gemini, Azure or Bedrock) and decide which models each team may use
- Connect each application you rely on once; the connection is then available to every flow, agent and teammate
- Open Chat and describe the outcome you want in plain words, and let the platform assemble the first flow for you
- Or start from the builder: pick a trigger, a webhook, an app event, a schedule or a manual run, then add the steps
- Test each step on real data and read its output before you publish anything
- Drop into an inline TypeScript step, or the HTTP piece, for anything the ready-made integrations do not cover
- For an agent, describe the goal, pick the tools it may use, lock the input values it must not change and cap its steps per run
- Mark the steps that touch money, customers or production so they wait for a human approval
- Watch the run history to see what happened and retry the step that failed, and connect an MCP client if you want to trigger everything from your assistant
Pros & Cons
Pros
- The core is MIT licensed and genuinely readable, forkable and self-hostable, with hundreds of outside contributors
- The same enterprise build runs on the EU cloud, on your own infrastructure or fully air-gapped, at the same price
- Customer automations, runs, documents and prompts are never used as training data, stated without hedging
- Credentials are never returned by the API, never shown again and never exported, and can stay in your own vault
- Pricing is charged per run rather than per step, so a long flow costs the same single credit as a short one
- Flows are unlimited on every plan, including the permanently free one, which needs no credit card
- Bringing your own model keys brings every AI step back down to one credit and moves model cost to your provider bill
Cons
- The security page advertises SOC 2 Type 2 while the pricing FAQ says the SOC 2 programme is still in progress, a contradiction the buyer has to resolve
- The Terms of Service and Privacy Policy pages were both unavailable on the site at the time of review
- There is no mobile application and no browser extension
- AI steps cost 2, 10 or 20 credits depending on the model class unless you supply your own provider keys
- SSO and standard roles start at the Team plan, and SCIM, custom RBAC, secret managers, audit logs and event streaming are reserved for the quote-only Ultimate plan
- Human support by email begins at the Team plan; the Free and Plus plans rely on the community forum
- The sub-processor list and the DPA are sent on request rather than published, and no list of supported interface languages is available
Pricing & Plans
Activepieces offers a permanently free plan that requires no credit card and has no time limit, capped at 100 credits per day with unlimited flows. The cheapest paid entry point is the Plus plan at 16.00 USD per month when billed annually, which includes 10,000 credits per month and up to five users. Team costs 166.00 USD per month billed annually, and the Ultimate plan is quoted individually.
- 0 USD
- free forever with no card
- 100 credits a day
- unlimited flows
- Flows
- Agents and Tables
- MCP and API access
- 16 USD per month billed yearly
- 10
- 000 credits a month then pay as you go at 0.007 USD per credit
- up to 5 users
- bring your own AI keys
- 166 USD per month billed yearly
- 50
- 000 credits a month
- unlimited projects
- 25 users included
- SSO and standard roles
- global connections
- email support
- custom pricing
- tailored yearly credits
- dedicated workers
- execution priority
- SCIM and custom roles
- releases and Git Sync
- audit logs and secret managers
- dedicated support
- separate enterprise offering for putting the builder inside your own product
- priced on request
Data, GDPR & hosting
A consolidated view of how Activepieces handles your data.
GDPR overview
GDPR is addressed concretely rather than as a badge. The privacy policy last updated on 19 January 2026 cites Article 6.1 (c), (d) and (f) as the lawful grounds for processing data of people in the EEA, and opens a dedicated section for EU, UK and EEA residents covering withdrawal of consent, the right to be forgotten and data portability. Requests go to privacy@activepieces.com, and personal data is deleted within twelve months of an account being closed. Transfers to third parties are framed by Data Processing Addenda, EU Model Clauses or Binding Corporate Rules, and the pricing FAQ says a DPA and a current sub-processor list are sent on request. Two gaps remain: no Article 27 EU representative is named, and no data protection officer is identified anywhere on the site.
Who owns the data?
The customer keeps ownership of what they put into the platform. Automations are plain JSON and export through the API together with tables, files and run history, and because the core is MIT licensed the same automations can keep running on the customer's own servers after leaving. Credentials are the strictest case: a saved connection is never returned by the API, never shown again in the builder and never included in an export, and it can be kept in the customer's own vault so the vendor never stores it at all. The privacy policy states that Activepieces neither sells nor rents customer data. The agreement is governed by the law of Delaware.
Reuse rights
Customers may use, export and reuse their own content freely, without asking permission: automations are exported as plain JSON through the API, with tables, files and run history, and the MIT-licensed core can be redeployed elsewhere and run the very same automations. The vendor's own use is narrow. Personal data is processed to run the service, take payments, provide support and detect fraud, with GDPR Article 6.1 grounds cited for people in the EEA. Customer automations, runs, documents and prompts are never used as training data, for Activepieces or for anyone else, and Google Workspace data is excluded from advertising, profiling and model training. Named third parties handling data include Google Analytics, Intercom, Mixpanel, Userback and Stripe.
Data retention & training
Hosting summary
The Activepieces cloud runs in Frankfurt, Germany by default, and the vendor says another region can be arranged on request. Self-hosting is a first-class alternative rather than a fallback: the same enterprise build runs on your own Docker or Kubernetes infrastructure, in the region your policy requires, on hardware nobody else shares and as a single tenant. Beyond that, an air-gapped deployment keeps everything inside your network, with the model answering your agents running on your own machines and no outbound traffic. Data is encrypted with AES-256 at rest and TLS in transit, and every credential is encrypted as its own value. Two nuances belong on the record: the privacy policy states that personal data provided directly to Activepieces is transmitted to the United States to fulfil the contract, and card payments are processed by Stripe in the United States. The resolved web server address is a Cloudflare anycast node and says nothing about where customer data actually lives.
Things to keep in mind
Risks and trade-offs to weigh before adopting Activepieces.
- An agent acting across live business systems can move money, message customers or change production records; the approval gates exist for a reason and switching them off to go faster is the main operational risk
- The certification claim is inconsistent between the security page and the pricing FAQ, so do not treat SOC 2 Type 2 as established without asking for the report
- The Terms of Service and Privacy Policy were unavailable on the site at review time, meaning you may be committing to terms you have not been able to read
- Credit consumption is easy to underestimate once agents and AI steps are involved, since a single frontier-model step costs twenty times a plain run
- Delegating recurring judgement to agents erodes the team's own grip on how the work is actually done, and the knowledge of why a process exists fades with the people who automated it
- Self-hosting shifts responsibility for patching, backups and incident response onto your own team, and community-maintained deployment templates can lag behind the core release
- The sub-processor list and the DPA are only sent on request, so a data protection assessment cannot be completed from the public site alone
Setup & Integrations
Technical difficulty
Low on the cloud: you sign up without a card, connect an application once, and Chat can build a first flow from a sentence with no prompt engineering. Self-hosting is a moderate step, a single Docker command or a Kubernetes deployment, with third-party templates for AWS, GCP, Railway and others that are maintained outside the core team and may lag. The genuinely technical work is administrative rather than developmental: wiring SSO and SCIM, connecting an external secret manager and streaming audit logs to a SIEM all require someone on the IT side.
Deployment
Integrations
Supported languages
Behind Activepieces
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Activepieces.
Frequently asked questions
Do I need to know how to code to use Activepieces?
Can I run Activepieces on my own servers?
Is Activepieces really open source?
How does the pricing actually work?
Are my data or my prompts used to train models?
Can an agent see my credentials?
Is there an API, and is there a mobile app?
What happens to my automations if I leave?
Should you pick Activepieces?
Activepieces makes an unusually coherent argument: an automation and AI-agent platform whose core is genuinely open source, that runs identically on a European cloud, on your own cluster or inside an air-gapped network, and that charges per run rather than per step. For a company that wants to give business teams real automation without handing an outside vendor its credentials, its model choice or its data residency, the combination is hard to find elsewhere. The governance layer is not an afterthought either: sealed projects, SSO and SCIM, action-level permissions, audit trails that stream to a SIEM, and secrets that can stay in your own vault.
Two reservations deserve to be stated plainly. First, the compliance story does not hold together on the site itself: the security page advertises SOC 2 Type 2 while the pricing FAQ says the SOC 2 programme is still in progress. Anyone buying under a regulated procurement process should get that settled in writing before signing. Second, at the time of this review both the Terms of Service and the Privacy Policy pages returned "This document is not available right now", which is an awkward gap for a vendor selling on trust and control.
The commercial shape is also worth understanding before committing. The free plan is real and permanent, but the governance features that justify the platform for an enterprise start at the Team plan and, for the most demanding ones, at the quote-only Ultimate plan. Self-hosting the Community Edition gives you the automation core, not Agents, Chat, projects or audit logs. Taken together, Activepieces is a strong candidate for teams who value portability and control above all, provided they check the compliance documentation themselves.
- Choosing a selection results in a full page refresh.
- Opens in a new window.