A reasoning layer for AI-native product managers.
Every AI session starts cold. You re-explain the stakeholder context, re-derive the decision, hold the thread across PRDs, standups, and 1:1s in your head until you don't. What if every session built on the last, and the system held the thread for you?
OS-Intelligence is a free, open-source context system for product managers using Claude Code. It captures your meetings, PRDs, stakeholder notes, and chats, synthesises them into reference state, and surfaces the right context at the start of every session. It's the rails for AI-native PM work: every interaction has a reason, every decision is traceable, the system shows its homework.
Built by Simon Conway, an AI-native PM. In daily use, designed to be shaped by the PMs who use it. If you live in Claude Code and you've already started building your own context system to make it work, this is for you.
Files stay on your machine. The only network call is the one Claude Code already makes to Anthropic.
New to Claude Code? Start with the Setup guide. It walks you through installing Claude Code, Cursor, and OS-Intelligence in about 15 minutes. No coding skills needed.
Already have Claude Code installed? Skip to Get started below.
In your terminal:
mkdir -p ~/Code && cd ~/Code # or any other location
git clone https://github.com/simonconway1979/os-intelligence.git
cd os-intelligence
claudeOnce Claude is running, you have two ways in.
Quick look (~60 seconds) — open the Acme Example. A fully synthetic PM project with 30+ docs already synthesised. Walk around inside a populated system before committing your own data.
/os-start
Pick "Acme Example" from the menu.
Guided walkthrough (~20 minutes) — /os-welcome sets up your first project with your own data and walks you through the full loop.
/os-welcome
OS-Intelligence is the framework. Your meetings, documents, notes, and people files are the context, and that context is private. Don't fork this repo on GitHub. Forks default to public, so anything you add gets indexed.
Recommended: create your own private repo and repoint the remote so commits go there. Full walkthrough in the Setup guide.
OS-Intelligence has its own vocabulary. Worth a quick pass before you dig in:
intelligence/— your project's source-of-truth folder. Meetings, docs, chats, notes all live here. Synthesis reads from this.current-state.md— the living context summary for a project. Updated by/os-save,/ctx-doc,/ctx-synthesise. Read at the start of every session./os-*skills — operating the system itself. Start a session, save a session, switch project, add a project./ctx-*skills — adding to context. Documents, transcripts, notes, chat threads, cross-synthesis.- Project vs portfolio — a project is one initiative; a portfolio holds many items of the same shape (ideas, opportunities, events).
- "Reasoning layer" — the system captures, structures, and surfaces context. Claude does the language work on top.
The value of OS-Intelligence increases as your context layer develops. This happens naturally as you use it.
After the /os-welcome walkthrough you'll have a feel for how it works and where it could fit. A few weeks in, the system is doing real lifting on the things you used to re-explain. Past that, it's hard to remember how you worked without it.
You can see a fully populated OS-Intelligence project at projects/acme-corp-example/ — synthesised context across docs, meetings, chats, and people for a (fully synthetic) scenario. To get started, take a look at the example project tour.
Each skill is a lightweight workflow that helps you manage people, projects, and context.
See the Skills reference for what each skill does, grouped by capture, synthesise, retrieve, and scaffold.
Set up your status bar to show your current model, context usage, session cost, and anything else you'd otherwise switch tabs to check. Stops you from breaking flow to ask Claude what model you're on.
See the Status bar guide. Adapted from Hannah Stulberg's article on the same.
How the pieces fit, projects vs portfolios, the context flow: ARCHITECTURE.md.
OS-Intelligence is the reasoning layer. It expects you'll bring your own skills, frameworks, and workflows on top.
- Tone of voice ·
context-library/tone-of-voice.md. Default styles inwriting-styles-global/. Edit them. - MCP setup ·
context-library/mcp-setup.md. Paste your own routing. - Add-ons ·
docs/addons.md. Where to drop your own skills, sub-agents, and frameworks.
If you already have a working CLAUDE.md, keep it. Just preserve the stakeholder-intelligence section so the ctx- skills know where to write. See CLAUDE.md.
A local activity log at .os-intel-log/usage.jsonl records skill usage and session counts so you can see your own patterns. It never leaves your machine. See docs/usage-log.md to inspect or disable.
Apache 2.0 with a contributor licence agreement. See CONTRIBUTING.md, CODE_OF_CONDUCT.md, AUTHORS.md.
In active development. Open an issue or reach out directly.
Apache 2.0. See LICENSE.
OS-Intelligence is built by Simon Conway, an AI-native product manager.
20 years in product across SaaS, consultancy, and operating-model transformation. Most recently led an AI-transformation North Star at a PE-backed enterprise software unicorn, launching it to a 1,000-person All Hands and aligning a 2,000-person R&D organisation. Earlier: Product Principal at AND Digital (Constellation Brands operating-model transformation, tier-one bank mobile app at 5M users), product and AI research at Capgemini, and 14 years as Director of Product at Alps Education building 0-1 products deployed into ~2,000 UK schools.
Now operating independently. Claude Code is the daily driver. OS-Intelligence is the public, free version of the system Simon uses every day. The fiddly migration (overlaying it onto an existing PM setup so you keep your workflows and gain persistent context) is also offered as paid consulting.
Reach out via LinkedIn or open an issue on this repo.