Your growth data, on your own machine — ask it anything in plain English.

Infinite OS is a self-hosted, local-first growth-analytics runtime. It connects your data sources — Google Analytics 4, PostHog, Stripe, Meta, Shopify, and read-only X public post metrics — into a Postgres database on your own machine, keeps it synced, and lets you ask questions like "how many page views in the last 7 days?" or "which channels drove the most traffic?" and get real answers from governed, source-accurate metrics.

🌐 Site: infinite.fast   ·   𝕏 Project: @InfiniteOS_   ·   Maker: @RiverKhan

It is not a broad agent runtime, skill platform, or generic SQL tool. For natural-language questions it forwards your prompt to your own Claude or Codex account using your own credentials; it executes no tools or shell commands. Your data never leaves your machine unless you send it somewhere.

How it works

Everything runs locally as one small stack:

• Postgres — your synced growth data, sync state, jobs, schedules, and queryable views (a rebuildable cache; the source of truth stays with the providers).
• app daemon — a local HTTP API the CLI talks to.
• worker — owns the scheduler and runs syncs as background jobs.
• CLI (infinite) — the operator shell + the chat interface.
• encrypted connector credentials — provider keys/tokens are encrypted at rest with your GROWTH_OS_ENCRYPTION_KEY.

You connect a source once, sync it, then ask questions. Answers come from a typed metric layer with authority/provenance rules — not free-form SQL — so a number is either source-accurate or honestly reported as unavailable.

Example questions

Once a source is connected and synced, just ask — in plain English:

• "How many page views did I get in the last 7 days?"
• "What were my top pages this week?"
• "Which channels drove the most traffic last month?"
• "How many new users this week vs last week?"
• "What's my engagement rate over the last 30 days?"
• "How much revenue did I make in the last 30 days?" (Stripe)
• "Compare this week's traffic to the previous week."

Infinite figures out the right metric, runs it against your synced data, and answers with the numbers + a short read — citing the source and flagging anything it can't verify (it won't guess).

Quickstart

Install with one command (macOS / Linux). It checks you have git, Node ≥ 20, and pnpm (it won't install or change them for you), puts Infinite at ~/.infinite/app, drops an infinite command on your PATH, and runs setup:

curl -fsSL https://raw.githubusercontent.com/Infinite-Labs-AI/infinite-os/main/scripts/install.sh | bash

Then:

infinite setup     # connect a data source + configure the model, and start the local stack
infinite           # ask a question
infinite "how many page views in the last 7 days"

From source

git clone https://github.com/Infinite-Labs-AI/infinite-os.git infinite
cd infinite
./infinite setup
./infinite

setup installs/builds the workspace if needed, configures model auth, and starts the bundled Docker stack.

Commands

Connectors

Connectable: Google Analytics 4 · PostHog · Stripe · Meta · Shopify · X (read-only public post metrics). Deeper attribution and content analysis are on the roadmap.

Install the tracking tag on your site

GA4 and PostHog only start collecting data once their tracking tag is on your website. After infinite setup connects an analytics source, install the tag into your own site's repo with our published npm package, infinite-tag — it uses only your public keys, auto-detects your framework, and writes idempotent, fully reversible changes:

# run inside your website's code repo
npx infinite-tag@latest install

infinite setup prints a ready-to-paste npx infinite-tag install … command with your Measurement ID / PostHog key and workspace id already filled in (and saves your public keys to ~/.infinite/artifacts/, so a bare npx infinite-tag install discovers them automatically). See packages/instrument/README.md for all flags and the supported frameworks (Next.js, Vite + React, static HTML).

Configuration & data safety

infinite setup writes your config and secrets into a gitignored .growth-os/ directory; connector credentials are encrypted at rest. Your data stays on your machine — nothing is sent anywhere unless you do it. See SECURITY.md for the trust model and the full variable list.

Develop

pnpm install
pnpm typecheck
pnpm test

See docs/local-and-docker-quickstart.md for the Docker path, CONTRIBUTING.md to contribute, and SECURITY.md to report a vulnerability.

License

MIT © Ultima AI, Inc.