Blog: Introducing MCP Integration — Extend SaaS Factory Agents With Any Tool

Introducing MCP Integration — Extend SaaS Factory Agents With Any Tool

v1.0.161 — March 2026

Since day one, SaaS Factory's agents have been opinionated about their toolchain: GitHub for code, Anthropic for reasoning, Vercel for deployment, Neon for data, Stripe for payments. That stack works extremely well for the products we build. But it isn't the only stack your business runs on.

Today, we're changing that.

SaaS Factory v1.0.161 introduces MCP (Model Context Protocol) support and a Plugin Marketplace — making our agent toolchain fully extensible for the first time.

The Problem With Hardcoded Tools

Every capability our agents had lived inside a single file. 41 kilobytes of tool implementations, all hand-wired to the same five services. If you ran Datadog for observability instead of something we'd already integrated, agents couldn't see it. If your team tracked work in Linear, agents didn't know. If your data lived in Snowflake, agents couldn't query it.

The workaround was always the same: open a support request, wait for us to add an integration, get a new platform release. That's not autonomous. That's a bottleneck wearing a different hat.

What MCP Changes

Model Context Protocol is an open standard — backed by Anthropic and rapidly adopted across the AI tooling ecosystem — that defines a common language for AI agents and external tools to communicate. Think of it as USB-C for agent capabilities: any compliant tool can plug in, and the agent knows how to use it.

With MCP support in SaaS Factory:

200+ services are available immediately in the Plugin Marketplace, including Asana, Zapier, Confluence, Datadog, Sentry, Snowflake, Linear, Notion, and many more.
Your internal APIs can be connected too. If your team has built internal tooling, expose it as an MCP server and agents will use it just like any other tool — no platform PR required.
Tool registration is dynamic. Agents discover available tools at runtime from a live registry. Adding a new tool doesn't require touching any platform code.
Configuration is per-workspace. Different teams can run different tool sets without stepping on each other.

How It Works

The MCP bridge sits between our agent execution layer and the outside world. When an agent begins a task, it queries the MCP registry for available tools in the current workspace. Those tools — whether they're Zapier automations, Datadog dashboards, or your company's internal ticketing API — are presented to the agent the same way built-in tools are. The agent decides what's relevant and calls what it needs.

Our core integrations (GitHub, Anthropic, Vercel, Neon, Stripe) remain first-class. MCP doesn't replace them — it extends everything around them.

Getting Started

Plugin Marketplace is live now for all workspaces. Head to Settings → Integrations → Plugin Marketplace, find the tools your team uses, and connect them. For custom internal tools, the Add Custom Server flow accepts any MCP-compliant endpoint.

Once a plugin is connected, your agents will start using it on their next task run. No agent retraining, no configuration files, no restarts.

What This Means for the Platform

SaaS Factory has always been model-agnostic and infrastructure-agnostic in principle. With this release, we're tool-agnostic in practice.

As the MCP ecosystem grows — and it's growing fast, with Cursor, Devin, and a widening list of AI development tools standardizing on it — every new MCP server that ships anywhere becomes a new capability your Factory agents can use. The platform gets smarter not just when we ship updates, but whenever the ecosystem does.

That's the compounding loop we've been building toward.

Full documentation for MCP integration and the Plugin Marketplace is available in the MCP Server Integration & Plugin Marketplace guide.