Activepieces
Open source, MIT licensed automation platform where every one of 700 plus integrations is also an MCP server, combining a visual flow builder, AI agents and tables with cloud or fully self hosted deployment.
Activepieces is an open-source, AI-first automation platform, positioned as an alternative to tools like Zapier, Make, and n8n but built around artificial intelligence and the Model Context Protocol rather than treating them as add-ons. Teams use a no-code visual builder to assemble workflows from reusable building blocks the platform calls pieces, each one an integration such as Gmail, Slack, or a database, connected into flows that start from a trigger like an incoming email or a schedule and run a series of actions, with branching and data transformations along the way.
The pieces themselves are written in TypeScript and published as open-source packages, and a large share of them come from the community rather than the core team, which keeps the integration library growing and fully inspectable. On top of this foundation, Activepieces adds genuine AI capabilities: native steps that call language models, an AI assistant that helps build flows, and AI agents that can sense their environment, plan a sequence of steps, and act by using pieces as tools, moving a workflow from fixed rules toward something more autonomous.
Its most distinctive bet is on the Model Context Protocol. Every integration in Activepieces is automatically exposed as an MCP server, which turns its whole library into tools that external AI assistants like Claude, Cursor, or other MCP-compatible clients can call directly, so an AI agent can send an email, update a CRM, or edit a spreadsheet through Activepieces without a custom workflow for each case. The company describes this as one of the largest open-source MCP toolkits available.
Because it is open source under a permissive license, Activepieces can be self-hosted entirely within an organization's own network, including air-gapped environments, which appeals to teams with strict requirements, and it is also offered as a managed cloud with regional data options. Human oversight is treated as a first-class part of the design, with approval and delay steps and audit trails, and enterprise editions add governance features like single sign-on, user provisioning, and role-based access. The result is a platform that suits both non-technical builders and developers who want to extend it.
Vendor details
Canonical URL
https://www.activepieces.com
Category
Agent builder
Funding status
Independent. Y Combinator backed, founded 2022. Open source core under the MIT license with more than 700 integration pieces, roughly 60 percent community contributed, and over 300,000 builders reported. Customers named on the site include MoneyGram, Alan, PostHog, Red Bull, Rakuten, Contentful, Roblox, DocuSign and Pipedrive.
Company status
independent
Use cases & customers
Target customers
Deployment options
Integrations
More than 700 open source TypeScript integration pieces spanning CRM, communication, databases, project management, ecommerce, HR and AI providers, each usable as a workflow step or as an agent tool, with anything else reachable through the HTTP piece or a webhook. Every piece is automatically exposed as an MCP server callable from Claude, ChatGPT, Cursor, GitHub Copilot, Windsurf and Gemini, which the vendor describes as the largest open source MCP toolkit. An AI provider gateway connects Azure OpenAI, Amazon Bedrock, OpenRouter or LiteLLM, and an embeddable JavaScript SDK ships automation inside the customer's own product.
In practice
You want to move off expensive per-task automation and keep your data in-house. Activepieces is open source and self-hostable, even air-gapped, and replaces Zapier-style workflows while letting you inspect and extend every integration.
You want your AI assistant to actually do things in your apps. Activepieces exposes every integration as an MCP server, so Claude or Cursor can send email, update your CRM, or edit sheets through it directly.
Your workflow needs to make judgment calls, not just follow fixed rules. Activepieces lets you drop AI agents into a flow that sense, plan, and act using your connected pieces, with approval steps to keep humans in control.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
11.5 / 14 capabilities · 82%
| Integrations & Tool Calling | Full |
|---|---|
|
More than 700 integrations spanning CRM, communication, databases, project management, ecommerce, HR and AI providers, each a typed TypeScript piece usable as a workflow step or as an agent tool, with anything else reachable through the HTTP piece or a webhook; roughly 60 percent are community contributed and all are open source npm packages. Sourceactivepieces.com/pieces and github.com/activepiecesread 2026-08-30 |
|
| Workflow Orchestration | Full |
|
A visual builder composes multi step flows with branching, loops, conditions, code steps and data transformations, and AI agents can sense, plan and act using pieces as tools within those flows; agents, flows and tables ship on every plan including free, and the pricing page documents support for multi agent orchestration at scale. Sourceactivepieces.com/pricing and product/ai-agent-builder; activepieces.com/product/ai-agent-builderread 2026-08-30 |
|
| Knowledge Grounding & RAG | Full |
|
Each agent has a Knowledge Base section: upload documents (PDF, DOCX, TXT or CSV) or point at a project table, and the agent looks answers up instead of guessing; retrieval runs on Postgres with the pgvector extension, installed on Cloud and the standard self-hosted images, and table sources stay current as flows update them. Sourceactivepieces.com/docs/agents/knowledge and /docs/tables/with-agentsread 2026-09-29 |
|
| Human Oversight & Guardrails | Full |
|
Human in the loop ships as first class flow steps: require approval before continuing and delay execution for a period, alongside chat and form input interfaces; the governance layer adds sensitive connections that require approval before a user may use them, and flows are fully versioned so changes can be reviewed and rolled back. Sourcegithub.com/activepieces and activepieces.com/product/governance-and-managementread 2026-08-30 |
|
| Security, Identity & Governance | Full |
|
Deep named control surface documented on the governance page: SAML 2.0 single sign on with Okta, Microsoft Entra ID, Google Workspace, Auth0 and OneLogin, SCIM provisioning, advanced role based access control, audit logs with full actor, resource and timestamp context, sensitive connections requiring approval before use, centralized credentials, private pieces, per team integration visibility control, secret managers and event streaming to a customer SIEM. Sourceactivepieces.com/product/governance-and-management and pricingread 2026-08-30 |
|
| Observability & Auditability | Full |
|
Audit logs record every flow run, user action and system event with resource type, detailed description, actor and timestamp, filterable by action, performer, project and date range, with event streaming to a customer SIEM, team analytics, and per step run logs exposing inputs, outputs and errors; flows are fully versioned. Sourceactivepieces.com/product/governance-and-management and pricingread 2026-08-30 |
|
| Memory & State Persistence | Partial |
|
Tables store data inside Activepieces for flows and agents to read and write, so state can persist between runs without an external database, and flows are versioned. A table is storage the builder's flows and agents read and write rather than agent memory, and no agent memory with a stated scope, lifetime or review path is documented in the agent docs. Sourceactivepieces.com/docs/tables/overview and /docs/agents/createread 2026-09-29 |
|
| Deployment & Data Residency | Full |
|
Every plan runs on the managed cloud or on the customer's own servers, with the MIT licensed Community Edition self hostable via Docker, Docker Compose, Kubernetes or Helm on any cloud including air gapped networks, and the managed cloud offering EU and US data regions plus dedicated workers on Ultimate. Sourceactivepieces.com/pricing, product/deployment-options and open-source; activepieces.com/product/deployment-optionsread 2026-08-30 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
A Template Library in the Explore tab offers ready-made flows across Sales, Marketing, Customer Service, Finance, HR, Engineering, Operations and AI Agents, searchable by name and showing the apps used and estimated time and cost saved; admins can manage templates for their platform. The docs name categories but no individual template or ready agent. Sourceactivepieces.com/docs/flows/template-library and /docs/admin-guide/guides/templatesread 2026-09-29 |
|
| Triggers & Channel Coverage | Full |
|
Triggers span app events across more than 700 integrations, scheduled runs, webhooks and HTTP for anything with an API, plus built in human input triggers including a chat interface and a form interface, and invocation from external AI clients including Claude, ChatGPT, Cursor, GitHub Copilot and Gemini through the MCP surface on every plan. Sourceactivepieces.com/pricing and github.com/activepiecesread 2026-08-30 |
|
| Model Flexibility & Routing | Full |
|
The governance page documents an AI Providers surface where customers connect their own AI gateway or cloud provider, naming Azure OpenAI, Amazon Bedrock, OpenRouter and LiteLLM with centralized billing and usage; the pricing page adds bring your own AI keys from the Plus plan and customer selection between fast, smart and frontier model tiers priced at 2, 10 and 20 credits. Sourceactivepieces.com/product/governance-and-management and pricingread 2026-08-30 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Every one of the 700 plus pieces is automatically exposed as an MCP server callable from Claude, ChatGPT, Cursor, GitHub Copilot, Windsurf and Gemini, described by the vendor as the largest open source MCP toolkit; MCPs and API access ship on every plan including free, the piece framework is open source TypeScript published to npm with hot reloading, and an embeddable JavaScript SDK is offered for SaaS integration. Sourceactivepieces.com/pricing and github.com/activepiecesread 2026-08-30 |
|
| Testing, Debugging & Optimization | Partial |
|
Individual steps can be run in a test mode with sample data before a flow is published, run logs expose inputs, outputs and errors per step, failed steps retry automatically, and versioning with Git Sync and releases on Ultimate allows changes to be shipped and rolled back; no agent evaluation, scoring or regression testing product is documented. Sourceactivepieces.com/docs and pricingread 2026-08-30 |
|
| Browser & Computer Use | Not documented |
|
All action is taken through typed integration pieces, HTTP requests and webhooks against documented APIs; no browser automation, screen control or ability to operate software lacking a programmatic interface is documented on the product pages or in the open source repository. Sourceactivepieces.comread 2026-08-30 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Recent platform changes
Activepieces 0.92.0 adds agent creation from a natural language description and an AI Router that chooses a workflow route from a question. The release also adds a Deno code sandbox, an MCP activity feed recording client calls, and OpenID Connect support for restricting connector access by email domain.
Bears on: Agent capability
View sourceActivepieces release 0.85.4 added AI-ready metadata across hundreds of pieces, tagged non-agent actions with audience='human', and updated builder docs to require AI metadata on new actions and triggers.
Bears on: MCP / tool calling / API
View sourceActivepieces release 0.85.4 shipped broader runtime visibility improvements, including evlog-wide events replacing OpenTelemetry + pino, date ranges for queue health metrics, and worker/sandbox state and RAM visibility.
Bears on: Observability / auditability
View sourcePricing
Free forever with 1,000 credits a month · Plus $20 a month (10,000 credits, up to 5 users) · Team $200 a month (50,000 credits, 25 users) · annual billing saves two months · Ultimate and Enterprise Control custom · MIT self host free
Credit based: one credit per flow run regardless of step count, one per agentic action, and 2, 10 or 20 per AI step depending on model tier, or 1 with your own AI key. Plan sets seat count and feature access.
Included quota
Free: 100 credits a day refreshed each morning, 1 user, unlimited flows, agents, tables, MCPs and API access. Plus $16/mo: 10,000 credits a month, 5 users, bring your own AI keys. Team $166/mo: 50,000 credits a month, 25 users then $25 each, unlimited projects, SSO, standard roles, global connections, email support. Ultimate: custom credits pooled yearly, SCIM, custom RBAC, audit logs, secret managers, event streaming, private pieces, branding, Git Sync and releases, dedicated workers. Embed from $36k a year.
What is public
The pricing page lists Free ($0, 1,000 credits a month, one user), Plus ($20 a month, 10,000 credits, up to 5 users), Team ($200 a month, 50,000 credits, 25 users), Ultimate (custom, credits sized yearly, unlimited users), Embed (from $36k a year) and Enterprise Control (custom; self-hosted or air-gapped with SSO and SCIM, custom RBAC and audit logs). Extra credits are $0.007 on Plus and Team, and yearly billing saves two months.
Billing mechanics
Credit based: a monthly credit allowance per plan, with extra credits at $0.007 on Plus and Team and volume rates on Ultimate; bringing your own AI key lowers AI step cost. The self hosted Community Edition has no credit caps.
Cost watchouts
Governance is paywalled: SSO and standard roles start at Team, while SCIM, custom RBAC, audit logs, secret managers, event streaming and private pieces are Ultimate only, and the self hosted Community Edition excludes Agents and Chat, Projects, API access and the whole team and admin layer. Team includes 25 users then $25 per user per month. AI steps cost 2 to 20 credits depending on model tier unless you bring your own key. Self hosting shifts infrastructure, upgrades and uptime to you.
Variable cost rationale
Cost tracks credits consumed rather than seats, and one credit covers a whole flow run however many steps it has, so ordinary automation volume is cheap and predictable. AI steps are the real variable at 2 to 20 credits each, mitigated by bringing your own AI key at 1 credit.
Additional watchouts
Thinner integration catalog and template library than Zapier/Make/n8n (start-from-scratch more often); cloud execution slower than n8n; younger/smaller community (less Google signal for troubleshooting); enterprise governance is top-tier-only; pricing has shifted and varies across sources; AI-agent depth is newer than the incumbents'
Overage / add-ons
Past the plan allowance, extra credits cost $0.007 each on Plus and Team; on Free, runs wait until credits refresh the next morning. Ultimate pools credits across the year. Self hosted Community Edition has no caps.
Sales call required
Mixed (some tiers require a call)
Free / trial
Free forever with 1,000 credits a month for one user, no card; the MIT licensed Community Edition self hosts free.
Lowest paid plan
Plus at $20 a month, or about $16.67 a month billed yearly
Commercial notes
Founded 2022 (Mohammad AbuAboud, Ashraf Samhouri), YC-backed; MIT-licensed (truly permissive - unlike n8n's Sustainable-Use/fair-code license); cleaner/simpler UI than n8n, closer to Zapier UX; AI-agent layer with native Model Context Protocol ('largest open-source MCP toolkit'); 22,000+ GitHub stars (smaller than n8n's 190,000+); ~4.8/5 on G2; positions as the open-source, scaling-friendly alternative to Zapier/Make; strong fit for privacy/compliance-conscious teams wanting self-hosting
Key ambiguities
Pricing has changed twice: the task based tiers and the later $5 per active flow model are both superseded by the current credit model verified on the vendor's pricing page 2026-08-30. Many third party sources still publish the older structures.
Cancellation / refund
Self-hosted CE is free (MIT, no contract); Cloud Standard is free for 10 flows, then self-serve $5/active flow/mo; Ultimate/Enterprise is custom-contracted; managed-host resellers (e.g., Elestio) bill infra hourly
Support SLA / resale
Community support (CE/free); priority support on higher tiers; Ultimate/Enterprise adds SSO/SAML, SCIM, custom RBAC, audit logs, custom AI model support, and dedicated support; SOC 2 Type II + GDPR/HIPAA on managed cloud (EU or US regions); embed SDK for white-label automation in your product; native MCP support
Missing data
Nothing material: the full rate card, credit costs per action type and the plan feature matrix are all published.
Related vendors
- Lindy — AI teammate that lives in Slack, answers from your tools and…
- AgentX — No code platform to build, evaluate, and deploy multi agent…
- AI Library — No-code platform for building and running AI agents, with a…
- Aigensei — Enterprise agentic workflow platform that runs business processes…
- Airia — Enterprise platform that unifies agent orchestration, security, and…
- Airtable — Relational app platform turned agent builder: Omni configures agents…
Alternatives to Activepieces
The closest documented capability profiles to Activepieces among agent builders tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Cohere North11.0 / 14A lighter documented profile than Activepieces
- Dify12.0 / 14Fuller documented coverage on Prebuilt Agents, Templates & Packs
- Joget11.0 / 14A lighter documented profile than Activepieces
- OutSystems12.5 / 14Fuller documented coverage on Prebuilt Agents, Templates & Packs and Testing, Debugging & Optimization
- Retool11.5 / 14Fuller documented coverage on Testing, Debugging & Optimization
- Superblocks10.5 / 14A lighter documented profile than Activepieces
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded