How It Works

No jargon. No architecture diagrams. Just the gist.

The short version

FoundersOS is a bridge between your AI and your business data. When your AI needs to look up a customer, check your finances, run a playbook, or recall a past decision, it reaches through that bridge and gets the answer. You just talk normally - the plumbing is invisible.

A little more detail

There’s a standard called the Model Context Protocol (MCP). Think of it like a USB port for AI. MCP lets AI tools plug into external data sources - databases, APIs, files, whatever - without needing custom code for each one.

FoundersOS is a set of MCP tools that connect to a Supabase database (basically PostgreSQL with a nice web interface). When you say something like “What’s on my plate today?”, here’s what happens:

1. Your AI reads your message and decides it needs to call the get_session_start tool.
2. It sends a request through MCP to FoundersOS.
3. FoundersOS queries your Supabase database for tasks, finances, CRM activity, and feed updates.
4. The results come back, and your AI formats a helpful answer.

The whole round trip takes about a second. You never see the database. You never write a query. You just ask.

Why Supabase?

A few reasons:

• Free tier that’s generous enough for most small businesses.
• Managed PostgreSQL - you don’t need to run a database server.
• pgvector extension for semantic memory search.
• Web dashboard so you can peek at your data anytime.
• Row Level Security if you need to lock things down for a team.

You own the database. Your data never passes through our servers. If you ever want to leave FoundersOS, your data is right there in standard PostgreSQL tables - export it, query it, do whatever you want with it.

Why MCP?

MCP is an open standard backed by Anthropic and adopted by a growing list of AI tools. By building on MCP instead of a proprietary API:

• You’re not locked into any single AI tool.
• You can use FoundersOS with Claude Desktop today and switch to something else tomorrow.
• Other developers can build compatible tools that work alongside ours.

What “open source” means here

The entire codebase is MIT-licensed. That means:

• You can read every line of code.
• You can modify it for your own needs.
• You can contribute improvements back to the community.
• You can use it commercially, no strings attached.
• Nobody can take it away from you.

There’s no “open core” catch, no premium tier waiting in the wings. The tools you see today are the tools you get. If we build something new, it goes in the same repo.

The tech stack, if you’re curious

• TypeScript - the MCP server and all tools
• Supabase / PostgreSQL - data storage
• pgvector - semantic search for the memory tools
• MCP SDK - the protocol layer
• Node.js - runtime

But you don’t need to know any of this to use it. That’s the whole point.