By msolecki
The executive team you can't afford yet — strategy, offer, pipeline, delivery, money and focus for a company of one.
Decides whether a plan survives contact with reality. Use to red-team a plan, audit its assumptions, or run a premortem before something becomes irreversible.
Decides what gets published and where, and owns how the founder sounds. Use for the content plan, drafting, audience research, and capturing the founder's voice.
Decides whether the company can afford something and whether it actually makes money. Use for the monthly close, runway, profitability, and rate raises. Gives no tax or legal advice.
Decides what deserves the founder's attention right now and which specialist handles it. Use for the daily brief, the weekly and monthly review, triage of a pile of obligations, the work queue, or when you don't know who to ask.
Decides what bet the company makes this quarter and what it kills. Use for quarterly planning, sizing a bet, kill-or-continue calls, and the annual review.
Read twelve months of decisions back and score the judgment rather than the outcome — run once a year, from decisions/ and nothing else
List what must be true for a plan to work, strike what is already evidence, and rank the rest by cost to test — run before the plan is expensive
Collect what the ICP actually says about their problem, verbatim and sourced — run before a content plan, and whenever the founder is about to write about a topic instead of a question
Decide whether a manual task is worth automating using payback arithmetic — run before building anything internal, and especially when the founder is enjoying it
Price a bet by what it costs when it is wrong and cap the downside in writing before it starts — run before any bet enters goals.md
Executes bash commands
Hook triggers when Bash tool is used
Modifies files
Hook triggers on file write and edit operations
Based on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Uses power tools
Uses power tools
Uses Bash, Write, or Edit tools
Uses Bash, Write, or Edit tools
A Claude Code plugin that gives a company of one — or a founder running several — an executive team: 13 agents, 49 skills, 10 scheduled cadences — strategy, offer, pipeline, delivery, money, focus and the cross-business portfolio, with one markdown workspace per business as shared state and a write-time ownership guard keeping the org from corrupting it.
Free and MIT-licensed. Founder OS runs inside your existing Claude Code environment and adds no account or subscription of its own.
Getting started · Example workspace · All 49 workflows · Product philosophy
This repository is both the plugin marketplace (install straight from it) and the source repo (validator, tests, design docs).
Product philosophy, the org chart, and what the plugin refuses to do:
founder-os/README.md. This file covers how the machine works and how to develop against it.
Before installing:
| Requirement | Purpose |
|---|---|
| Recent Claude Code | Founder OS is a plugin, not a standalone app. |
| Python 3 | Runs the ownership hook. |
| PyYAML | Enables the full ownership-map check; the hook degrades gracefully without it. |
cron (optional) | Runs scheduled cadences. Manual workflows do not need it. |
The 13 agents are specialized roles invoked when needed, not 13 autonomous processes running all day. Founder OS knows only what is recorded in its local Markdown workspace or supplied in the current session; it does not automatically sync a calendar, CRM, inbox, or bank account.
In Claude Code:
/plugin marketplace add msolecki/founder-os
/plugin install founder-os@founder-os
Then, once:
/founder-os-init # ~20 min onboarding, ends with your first brief
/setup-cadences # optional: cron entries so it runs without being asked
Scheduled jobs run only while that machine and its cron service are running.
See the complete
docs/getting-started.md guide before installing if
you want the requirements, data boundary, and first-week workflow in one place.
To see the output first, open the fictional but contract-shaped
examples/studio-north/ workspace and follow
q-0720a from the daily brief into the queue, quarterly bet, week, and review.
| Piece | Where | What it does |
|---|---|---|
| Agents | founder-os/agents/*.md | 13 role definitions. Frontmatter: name, description, skills[], tools: allowlist (+ Agent(...) edges for managers). Body: four mandated headings. |
| Skills | founder-os/skills/*/SKILL.md | 49 procedures. Role skills follow references/skill-template.md exactly; each declares its writes in metadata.writes. |
| Ownership map | founder-os/references/ownership.yaml | The single source of truth: workspace_files: (what init scaffolds), owns: (one owner per file), sections: (the headings each file may contain). |
| Write-time guard | founder-os/hooks/ownership-guard.py | A PreToolUse hook. Denies a subagent writing a file it doesn't own, and denies subagents any outbound-capable tool (Bash, WebFetch, mcp__*). Fails open, main thread always allowed. |
| Validator | scripts/validate_package.py | 14 build-time checks (below). CI runs it on every push. |
| Cadences | founder-os/skills/setup-cadences/SKILL.md | 10 cron lines on your machine calling skills headless — a plugin cannot ship a schedule, so this writes one, with your consent, once per business. Fences are slugged per business (# BEGIN founder-os:<slug>) so two businesses never clobber each other's schedule. |
| Multi-business | founder-os/references/multi-business.md | One workspace per business + a registry (~/.founder-os/businesses.yaml) + a portfolio workspace. The hook resolves all registered roots; context-load step 0 picks the business before any file opens. |
| Commands doc | founder-os/COMMANDS.md | Generated by scripts/generate_commands.py from the skills' own frontmatter — the user-facing catalogue, machine-derived so it cannot drift. CI fails if it is stale. |
npx claudepluginhub msolecki/founder-os --plugin founder-osUltra-compressed communication mode. Cuts 65% of output tokens (measured) while keeping full technical accuracy by speaking like a caveman.
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Memory compression system for Claude Code - persist context across sessions
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Unified capability management center for Skills, Agents, and Commands.
45% cost reduction measured. Cache expiry prevention, SubTask auto-delegation, zero-cost context restoration, real-time cost dashboard. The only Claude Code plugin built from CC source analysis.