Your AI product team in a single command.
From idea validation to launch — designed for solo founders who think in products, not code.
中文文档 · Architecture · Examples · Quick Start
Most AI coding tools assume you already know what to build. They're great at writing code — but they won't tell you if your idea is worth building, how to define the right scope, or when you're actually ready to ship.
For solo founders without a technical background, the hard part isn't the code. It's the decisions before the code.
Solo Product Agent fills that gap.
| Cursor / Copilot | ECC / Aider | Solo Product Agent | |
|---|---|---|---|
| Focus | Code generation | Code workflow | Product development |
| Target user | Developers | Developers | Non-technical founders |
| Starts from | A coding task | A coding task | An idea or problem |
| Validates market | ✗ | ✗ | ✓ Gate 0 |
| Defines product | ✗ | ✗ | ✓ Gate 1 |
| Designs UX | ✗ | ✗ | ✓ Gate 2 |
| Guards scope | ✗ | ✗ | ✓ Built-in |
| Post-launch loop | ✗ | ✗ | ✓ Compact iteration |
| Solo constraints | ✗ | ✗ | ✓ Every recommendation |
- Product-first, not code-first — starts with "is this worth building?" before touching any code
- Decision Gate model — only stops to ask you when a real decision is needed; AI handles the rest
- Adaptive depth — lightweight for weekend projects, thorough for funded ventures
- Mid-entry protocol — already have a PRD? Already launched? It meets you where you are
- Solo constraints built in — every recommendation asks "can ONE person actually do this?"
- Claude Code installed and authenticated
# Step 1: Register the marketplace
/plugin marketplace add zhoupppp/solo-product-agent
# Step 2: Install
/plugin install solo-product-agent@solo-product-agentJust describe your idea in plain language:
I want to build a tool that helps freelancers track which clients owe them money
The router detects your stage, assigns depth, and dispatches the right agent automatically.
Or use the explicit entry point:
/solo-product-agent:solo-router
[MID-ENTRY]
Project: My Invoice Tracker
Current state: I have a PRD and the backend is 60% done
Existing assets: PRD doc, Figma mockups, partial codebase
Goal: Prepare for launch
The router will locate your stage, identify gaps, and pick up from there.
┌──────────────┐ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐
│ Gate 0 │ │ Gate 1 │ │ Gate 2 │ │ Gate 3 │
│ Worth doing? │───▶│ What to build│───▶│ Good enough? │───▶│ Ready to ship│
│ Go/No-Go │ │ Scope Lock │ │ Ship/Iterate │ │ Ship/Hold │
└──────┬───────┘ └──────┬───────┘ └──────┬───────┘ └──────┬───────┘
│ │ │ │
Market Validator Product Definer Experience Architect Launch & Ops
MVP Selector Scope Guardian Dev Officer
The AI works autonomously between gates. You only get asked when a real decision is needed.
| Agent | Role | Activates at |
|---|---|---|
| Router | Detects stage, assigns depth, dispatches | Always — entry point |
| Market Validator | TAM/SAM/SOM, pain points, competitive analysis, PMF | Gate 0 |
| MVP Selector | Tech stack selection, build-vs-buy, minimum architecture | Gate 0→1 |
| Product Definer | Value proposition, business model, product skeleton | Gate 1 |
| Scope Guardian | PRD writing, scope defense, version roadmap | Gate 1 |
| Experience Architect | User flows, Aha moment, viral loop, UX contract | Gate 2 |
| Dev Officer | Dev oversight, traceability, functional + usability testing | Gate 2→3 |
| Launch & Ops | Release readiness, gradual rollout, data framework, feedback loop | Gate 3 |
| Depth | When to use | Time investment |
|---|---|---|
| Lightweight | Weekend project, internal tool, tech spike | Hours |
| Standard | Serious side project, small SaaS | Days |
| Full | Funded venture, high-stakes product | Weeks |
The router auto-detects depth based on signals (real money, external users, fundraising). You can always override.
After launch, new features use a compact 4-step cycle — not the full pipeline:
Signal detected → Quick validation → Define scope → Build → Ship → Observe → repeat
"I want to build a Chrome extension that blocks distracting sites during work hours."
- Gate 0: 10-minute quick check — is there demand? Any obvious blockers?
- Gate 1: Lean PRD — 3 core features, clear not-to-do list
- Gate 2: Basic flow walkthrough
- Gate 3: Direct launch, 3 key metrics
Typical timeline: 2-3 days
"I want to build a SaaS that helps freelancers manage client invoices and follow up on late payments."
- Gate 0: Market sizing + pain point map + competitive analysis + PMF judgment
- Gate 1: Value proposition + business model (three-pressure test) + full PRD
- Gate 2: Core user flows + Aha moment design + UX contract
- Gate 3: Staged rollout + North Star metric + feedback loop
Typical timeline: 2-3 weeks
"We raised pre-seed. We think there's an opportunity in B2B expense management for SMBs."
- Gate 0: Deep market validation with multi-source cross-verification + solo feasibility check
- Gate 1: Full product definition + competitive moat path + version roadmap
- Gate 2: Three user persona walkthroughs + formal user testing plan
- Gate 3: Complete gradual rollout + data dashboard + continuous feedback loop
Typical timeline: 4-6 weeks
5 inherited principles:
- Spec before execution — align on what to build before building
- Fact / inference / guess — AI never presents assumptions as conclusions
- Default reject, replace don't stack — adding one thing means cutting another
- Reverse self-check — actively find flaws, not just good news
- Contract traceability — every output traces back to an upstream decision
3 new efficiency principles: 6. Decision Gate driven — AI works autonomously; only stops for real decisions 7. Adaptive depth — higher risk = deeper analysis; lower risk = faster pass 8. Iteration first — post-launch improvements use compact loops, not full pipeline
Q: Do I need to know how to code? No. The agents guide you through product decisions in plain language. When it's time to build, they work alongside Claude Code to handle the technical side.
Q: Does this replace a developer? No — it replaces the product management, UX design, and project oversight work that a solo founder would otherwise do alone (or skip). You still need Claude Code (or a developer) to write the actual code.
Q: Can I use this with an existing project? Yes. Use the mid-entry template to describe your current state. The router will locate your stage and pick up from there without redoing completed work.
Q: How is this different from just asking Claude to help with my product? Asking Claude directly gives you one-off answers. Solo Product Agent gives you a structured workflow with built-in guardrails — scope defense, anti-drift rules, decision gates, and persistent state across sessions.
Q: What if I disagree with a Gate decision? You're always in control. Gates are recommendations, not locks. You can override any decision — the agents will note the override and continue.
Q: Is this only for software products? The framework works best for digital products (apps, SaaS, tools). The market validation and product definition agents can apply to any product, but the dev and launch agents are optimized for software.
solo-product-agent/
├── skills/ # 8 agent skill definitions
│ ├── solo-router/ # Entry point — stage detection & dispatch
│ ├── solo-market-validator/ # Gate 0: market sizing, pain points, PMF
│ ├── solo-mvp-selector/ # Gate 0→1: tech stack, minimum architecture
│ ├── solo-product-definer/ # Gate 1: value prop, business model
│ ├── solo-scope-guardian/ # Gate 1: PRD, scope defense, roadmap
│ ├── solo-experience-architect/ # Gate 2: flows, Aha, UX contract
│ ├── solo-dev-officer/ # Gate 2→3: oversight, testing
│ └── solo-launch-ops/ # Gate 3: release, rollout, data, feedback
├── docs/ # Deep-dive documentation
├── examples/ # Real-world usage examples
└── .claude-plugin/ # Plugin marketplace configuration
Contributions welcome. See CONTRIBUTING.md for guidelines.
Ways to contribute:
- Improve agent prompts based on real usage
- Add examples for new use cases
- Translate to additional languages
- Report issues with specific agent behaviors
If Solo Product Agent has been useful, a ⭐ helps others find it.
MIT — use it, fork it, build on it.
Built for Claude Code · Designed for the one-person company movement