Skip to content
OpenSourceSoftware.ai
Activepieces logo

Activepieces

activepieces · Workflow automation

GitHub stars

25,244
—Awaiting first sync

Awaiting first sync

No measured star history yet

What is Activepieces?

Activepieces aims to make automation useful for the whole organisation. Developers set up integrations in TypeScript, everyone else builds flows in a visual editor, and the same integrations double as tools for AI assistants.

Activepieces is an open-source automation platform and replacement for Zapier, with a no-code flow builder and 280+ integrations called pieces. Pieces are TypeScript npm packages and are also available as MCP servers for AI tools. The community edition is MIT-licensed and can be self-hosted.

Core capabilities

  • Build flows with loops, branches, retries, HTTP and code steps in a no-code builder.
  • Connect 280+ pieces, each also usable as an MCP server for AI assistants.
  • Add human approval steps, chat and form interfaces, and native AI pieces.

A builder for everyone

Flows support loops, branches, automatic retries, HTTP requests and code steps with npm packages, and are fully versioned. Templates and translations help non-technical users, while an Ask AI option in code steps helps them clean data without writing code.

  • No-code builder
  • Versioned flows
  • Templates

Pieces, MCP and AI

More than 280 open-source pieces are published to npm, around 60% of them contributed by the community, and each is available as an MCP server for AI tools. Native AI pieces and an AI SDK support building agents inside flows.

  • Integrations
  • MCP
  • AI agents

Human input and self-hosting

Flows can wait for a period of time or for approval, and chat and form interfaces collect input from people. Activepieces can be self-hosted and network-gapped; the community edition is MIT-licensed, with enterprise features under a commercial licence.

  • Human in the loop
  • Self-hosted
  • MIT

Where it fits

Use cases

  1. 01

    Automate a client's back office

    Connect the client's apps with pieces, add branches and retries, and hand them a no-code builder they can adjust themselves.

  2. 02

    Put a person in the loop

    Pause a flow until someone approves it, or collect input through a built-in form or chat interface before continuing.

  3. 03

    Give AI assistants real tools

    Expose the pieces you use as MCP servers so assistants such as Claude Desktop or Cursor can act on your systems.