Getting Started
Set up FoundersOS in about 5 minutes. A Supabase project, a few environment variables, and you're good to go.
You don’t need to be a developer to set this up. If you can copy and paste, you’re qualified. Let’s get your AI connected to your business.
New to working with AI tools? We recommend Read This First before you set up - five minutes on what these tools can and can’t do, and how to get real work out of them.
What you’ll need
Before we start, grab these few things:
- A Supabase account - Sign up free. The free tier is plenty to get going.
- An AI tool that supports MCP - Claude Desktop, Cursor, or anything MCP-compatible.
- Node.js installed - the config runs through
npx, which ships with Node. Node 22 is the minimum (a Supabase requirement), so grab the current LTS to be safe. If you don’t already have it, download it here. This is a one-time install. - An OpenAI API key (for the memory search feature) - Get one here. There are also other alternatives - see Embedding providers if you’d rather use AWS Bedrock or run everything locally - and free - with Ollama.
That’s it. No Docker, no servers, no DevOps.
Setup
Three things to do: create a database, run one SQL file, and connect your AI. The fastest path uses the setup wizard, which generates both the SQL file and the config for you, with everything matched up correctly.
Option A: Setup wizard (recommended)
-
Create a Supabase project
Head to supabase.com, create an account if you don’t have one, and spin up a new project. Pick any name and region. Grab the project URL and the secret key from your project settings - you’ll need them in a minute.
-
Fill in the wizard
Open the setup wizard at foundersmcp.com/setup. Enter your Supabase URL, your secret key, and your embedding credentials (OpenAI by default; see Embedding providers for the alternatives). Hit Generate. Your credentials never leave your browser - nothing is sent to our servers.
-
Run the database setup it gives you (Step 1 in the wizard output)
The wizard hands you a ready-to-run SQL file with your embedding dimension already filled in to match the provider you picked. Copy it, paste it into the SQL Editor in your Supabase dashboard, and hit Run. That one file sets up everything from scratch - extensions, all tables, indexes, security policies, and functions. When you see “Success. No rows returned”, your database is ready.
-
Paste the config into your AI client (Step 2 in the wizard output)
Drop the generated block into your client’s MCP config (
claude_desktop_config.jsonfor Claude Desktop, the MCP settings for Cursor, Cowork, Continue.dev, Zed, or any spec-compliant MCP client - the format is identical everywhere). -
Talk to your AI
Restart your AI tool and say “catch me up” or “what can you do?” Your AI will pull up the session dashboard or show you example workflows to try. On a fresh install, it offers to walk you through adding your first customer and task.
Option B: Manual configuration
-
Create a Supabase project
Same as step 1 above.
-
Run the database schema
In your Supabase dashboard, open the SQL Editor and run
setup.sqlfrom the repo.Dimension warning: the raw file ships with
vector(1024)(the AWS Bedrock default). If you’re using OpenAI’s default model (1536) or any other dimension, find and replace everyvector(1024)in the file with your dimension before running it. Getting this wrong breaks memory storage and search. The wizard does this substitution for you, which is why it’s the recommended path. More detail in Embedding providers. -
Connect to your AI
Add this block to your AI tool’s MCP configuration file. For Claude Desktop, that’s
claude_desktop_config.json; for Cursor, Cowork, and others, it’s the MCP servers section of their settings. The format is the same across clients:{ "mcpServers": { "founders-os": { "command": "npx", "args": ["-y", "@ourthinktank/founders-os@latest"], "env": { "SUPABASE_URL": "https://your-project.supabase.co", "SUPABASE_SECRET_KEY": "sb_secret_...", "FOUNDERS_OS_COMPANY_ID": "your-company", "FOUNDERS_OS_USER_ID": "your-name", "FOUNDERS_OS_TIMEZONE": "America/Los_Angeles", "EMBEDDING_PROVIDER": "openai", "EMBEDDING_MODEL": "text-embedding-3-small", "EMBEDDING_DIM": "1536", "OPENAI_API_KEY": "sk-..." } } } }Replace the placeholder values with your actual Supabase URL, secret key, and OpenAI API key.
Tip - staying up to date: The
@latesttag on the package name means that each time you restart your AI client, it pulls the newest published version automatically - the easiest way to stay current. Prefer to stay put? Drop the tag (just@ourthinktank/founders-os) or pin a specific version like@ourthinktank/founders-os@0.15.0(an illustrative version - pin whichever release you’re on), and you’ll only move when you choose to. -
Talk to your AI
Restart your AI tool and say “catch me up” or “what can you do?” On a fresh install, the AI will detect there’s no data yet and offer a guided walkthrough.
Your first session
On a fresh install, FoundersOS detects that your database is empty and offers to walk you through the basics. You can also explore on your own. Here are a few things to try:
What can you do?
Add Acme Corp as a new prospect
Create a task: send proposal to Acme, due Friday, high priority
Run the customer-onboarding playbook for Acme
What's my financial summary?
Remember: our pricing starts at $49/month for the starter plan
Import the starter RSS feeds
Catch me up
The “catch me up” command pulls a full session dashboard - tasks, finances, CRM activity, and feed updates in one shot. It’s the fastest way to orient yourself at the start of any session.
Environment variables explained
Here’s what each setting does, in plain English:
| Variable | What it does | Required? |
|---|---|---|
SUPABASE_URL | The web address of your Supabase project | Yes |
SUPABASE_SECRET_KEY | The secret key (starts with sb_secret_) that lets the tools read and write your data | Yes |
FOUNDERS_OS_COMPANY_ID | Your company identifier - scopes financial data and org memory | No (defaults to “default”) |
FOUNDERS_OS_USER_ID | Your name or handle - scopes personal memories and identifies you on the team | No (defaults to “default”) |
FOUNDERS_OS_TIMEZONE | Your timezone - used by date-aware tools and year-to-date math | No (defaults to UTC) |
EMBEDDING_PROVIDER | Which AI service generates memory search vectors (openai, bedrock, or ollama) | No (defaults to “openai”) |
EMBEDDING_MODEL | The embedding model to use - the provider default is used if omitted | No |
EMBEDDING_DIM | The dimension of the vectors (must match your model) | No (defaults to “1536”) |
OPENAI_API_KEY | Your OpenAI key for generating memory embeddings | Only if using OpenAI |
Tip: Don’t want to use OpenAI? You can use Ollama (completely free, runs on your machine) or AWS Bedrock instead. See Embedding providers for the specific config for each, a plain-language explanation of what embeddings do, and how to pick. The embedding dimension is permanent once set - changing providers later means re-embedding your memory table.
Running multiple companies
If you run more than one business (we see you, serial founders), just create separate MCP server entries with different FOUNDERS_OS_COMPANY_ID values. Each one gets its own clean set of books.
{
"mcpServers": {
"founders-os-companyA": {
"command": "npx",
"args": ["-y", "@ourthinktank/founders-os@latest"],
"env": {
"FOUNDERS_OS_COMPANY_ID": "company-a",
"...": "..."
}
},
"founders-os-companyB": {
"command": "npx",
"args": ["-y", "@ourthinktank/founders-os@latest"],
"env": {
"FOUNDERS_OS_COMPANY_ID": "company-b",
"...": "..."
}
}
}
}
What’s next?
Now that you’re set up, take a look at what each tool can do:
- CRM Tools - manage your customers and pipeline
- Task Tools - track work with dependencies, entity linking, and AI assignment
- Project Tools - first-class project records with task rollups
- Playbook Tools - reusable templates that spin up a whole project in one run
- Tag Tools - organize with conventions: #projects, @people, !states
- Financial Tools - track your money
- Memory Tools - give your AI a long-term memory
- RSS Tools - stay on top of the news
- Members + Restore - manage your team and recover deleted records
Or just start talking to your AI. Say “catch me up” for a full orientation, or “what can you do?” to browse example workflows.