Activepieces is an open source, AI-first automation platform that aims to be a friendly, extensible alternative to closed SaaS workflow tools. Its TypeScript-based "pieces" framework, hundreds of ready-made connectors, and recent Model Context Protocol (MCP) integration combine to make automations that are both no-code for users and deeply programmable for developers. You can explore the project on GitHub and its docs at activepieces.com/docs.
activepieces
Organization
activepieces
AI Agents & MCPs & AI Workflow Automation • (280+ MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI AgentsKey features at a glance
- Pieces framework (TypeScript): Build connectors and logic as npm packages with strong typing and hot reloading. See the repo root README.md and CONTRIBUTING.md.
- MCP integration: Expose flows as call-able tools for Claude Desktop, Cursor, and Windsurf. The dedicated piece sits in packages/pieces/community/mcp.
- Open ecosystem: Hundreds of connectors from GitHub and Slack to databases and AI providers. Browse the catalog.
Introduction to Activepieces
Teams hit a wall when stitching together cloud services: CRM updates, messages, spreadsheets, webhooks, databases, and increasingly AI models. Vendor lock-in, rate limits, and missing connectors slow everyone down.
Activepieces addresses this by shipping a full-stack, self-hostable workflow engine with a visual builder on top of a strongly typed integration SDK. Pieces live as versioned npm packages in the repository (see packages/pieces/community), and every contribution instantly expands the platform.
The twist: those pieces can also be exposed as MCP tools so AI coding environments like Claude Desktop, Cursor, and Windsurf can call your flows directly from chat.
Why MCP matters here
MCP, the Model Context Protocol, is a standard for connecting AI clients to external tools, data sources, and runtimes securely. Instead of hardcoding one-off integrations, an AI IDE can discover and invoke tools from an MCP server. Activepieces embraces MCP so your flow becomes a discoverable tool with typed parameters and a response channel, bridging no-code workflows with LLM-driven development (see MCP overview by Anthropic).
Why I like it
Activepieces fuses developer ergonomics (TypeScript, hot reloading, versioned packages) with a polished visual builder and enterprise-friendly deployment.
The MCP angle is practical: your automation shows up as a tool in your AI workspace, so you can ask "create a Google Sheet and fill it with this CSV" and the agent calls your flow using the same guardrails and credentials you manage in Activepieces.
Under the hood
The repository is a monorepo managed with Nx, with core packages for the engine, server, CLI, UI, and shared types. The notable directories include packages/engine, packages/server, packages/react-ui, and the ever-growing packages/pieces/community. The MCP piece defines a webhook trigger and a response action so flows can present a typed tool interface to clients.
import { createPiece, PieceAuth } from "@activepieces/pieces-framework";
import { mcpTool } from "./lib/triggers/mcp-tool";
import { replyToMcpClient } from "./lib/actions/reply-to-mcp-client";
export const mcp = createPiece({
displayName: "MCP",
auth: PieceAuth.None(),
actions: [replyToMcpClient],
triggers: [mcpTool]
});
The trigger, mcp-tool.ts, registers a tool with a name, description, and parameter schema; the action, reply-to-mcp-client.ts, sends a response back to the MCP caller, optionally continuing the flow. This small contract cleanly maps LLM tool-calling to a durable workflow run.
Use cases for MCP + Activepieces
Because every piece can be surfaced as an MCP tool, the catalog translates directly into AI-initiated workflows. A few grounded examples tied to the repo's existing pieces:
AI agent updates CRM records via HubSpot or Salesforce connectors.
Generate documents with PandaDoc or PDF tools and send via SendGrid.
Read/write structured data using Google Sheets, Airtable, or Supabase (recently expanded).
Orchestrate AI providers like OpenAI, Gemini, and Mistral from a single flow (OpenTelemetry instrumentation recently landed in core for deeper visibility (Activepieces, 2025)).
What's new: features and MCP-ready pieces
Recent release notes highlight steady progress across core and catalog (see Releases): new GPT-5 model support; an overhauled flow builder UI; OpenTelemetry support for the platform and BullMQ; enhanced agent builder v2; grouping MCP tools by piece name; dynamic properties resolved in MCP tools; and explicit fixes for "testing MCP triggers" in the builder. On the pieces side, new integrations like Aircall, Respond.io, Autocalls, Missive, Paperform, Softr, Podio, Bluesky, Hunter, Supabase triggers/actions, and more have been added. Because pieces are MCP-exposed, these translate into immediate MCP tools for AI clients.
Community and contribution path
The project runs a healthy contributor program and maintains a welcoming guide in CONTRIBUTING.md, with an active Discord. Many connectors are community-built and published to npm on merge. If you want to help, start with the developer docs (guide) and explore the MCP piece to understand the tool schema.
Potential PRs worth opening
MCP UX/docs: Expand the MCP piece README with usage walkthroughs for Claude Desktop and Cursor, including screenshots and example flows. Add a troubleshooting section (timeouts, parameter validation, long-running responses).
MCP developer experience: Add examples demonstrating streaming outputs and partial updates where applicable; provide JSON schema snippets and a typed helper for building inputSchema arrays in mcp-tool.ts.
Observability: Extend OpenTelemetry coverage with spans around piece execution and MCP response paths in packages/server and packages/engine.
New pieces: Propose connectors with clear APIs and public SDKs (for example, Supabase enhancements, Stripe features, or data sync utilities). Each addition becomes an MCP tool.
Tests and CI: Add targeted unit tests for MCP trigger parameter parsing and reply-to-mcp-client behavior; consider e2e flows that run with a mock MCP client in CI.
License and editions
Activepieces ships under a permissive dual structure: the Community Edition is MIT (see LICENSE), while enterprise-only packages live under packages/ee with a commercial license. This lets teams self-host and extend the core freely while adding enterprise features when needed.
About the company
Activepieces is an open source company focused on secure, AI-native automation for teams. It operates transparently in public, is active on Discord, and, per the website footer, is backed by Y Combinator. The product spans cloud and self-hosted options, with a growing set of AI capabilities like agents, copilot, and universal AI pieces documented across the site.
Closing thoughts
If you are looking for a workflow engine that plays nicely with AI tooling, Activepieces is one of the most practical bridges between everyday SaaS tasks and MCP-enabled IDEs. Start by exploring the repo, try the visual builder, and then expose a flow as an MCP tool to meet your AI agents where they work. Repository: activepieces/activepieces.
Activepieces: Open Source AI Automation Meets MCP