Blog · October 14, 2026
AI Isn’t an Add-On Here. It Runs Through All of Federated Core.
Every app built in, every connection scoped, and learning that never moves your data: how AI works across the whole Federated platform.
Most software adds AI the same way: a chat box in the corner of one app, connected to a model in someone else’s cloud, with your data flowing out to meet it. It helps a little inside that one app and nowhere else.
We took a different approach. In Federated Computer, AI isn’t a feature of one app. It’s part of Federated Core, the layer every app runs on. Here’s what that means in practice.
Every app speaks AI from day one
Every app in the Software Library ships with a built-in AI connector, using the Model Context Protocol (MCP). An AI model, whether Claude, ChatGPT, or a private model you run yourself, can read and act inside your apps right away. You don’t write custom code or hire consultants, and you don’t wait months for an integration.
And because Core ties your apps together, your AI can work across them. A customer emails a complaint into FreeScout. Your AI agent reads it, pulls the related invoice from Nextcloud, checks your team’s policy in Discourse, drafts a reply, and pings your team in Element. That’s one workflow across four apps, and none of it needs to be set up by hand.
Your data stays where it is
Agents fetch data live, straight from your own applications. Nothing gets copied into a third-party database somewhere on the internet. Your Federated install is single-tenant, so it belongs to you alone, and once you change your admin password, we can’t see inside it either.
Every connection is scoped, and you can switch it off
Each AI connection is locked down to what it needs. An agent set up for customer support can’t open your admin files. You can see exactly what every agent can reach, and you can switch off any single app’s AI access instantly without breaking anything else.
Smarter AI without sharing your data
Federated Core also supports federated learning. Your business trains a model locally, on your own private data, and only encrypted learnings (never the raw data) are shared with a central coordinator. Every participating business gets a smarter shared model, and nobody ever sees your files, emails, or customer records.
Room to grow
When you’re ready to run whole teams of agents next to your data, add the Agent Sandbox to the same install. A QA tester, a support triage agent, and a documentation writer can all run at once, each sealed in its own private space.
AI and MCP features are included on every plan, including our free self-hosted tier.