You just changed my priority assessment. Let me recalibrate.                                                                                                                                                                               
  The Dashboard Is Not Nice-to-Have — It's the Product                                                                                                                                                                                       
  I was wrong to put the dashboard onboarding in the "nice-to-have" bucket. Here's why your observation changes things:

  The CLI wizard serves developers who got blocked. They're technical, they'll type pip install proclaw, they'll figure it out.

  But the much larger audience is the people who can barely get OpenClaw installed in the first place. These are the users on Discord saying "how do I deploy this?" — they don't want a CLI. They want to click a button and have it work.
  MyClaw and Hostinger theoretically offered this but failed at marketing and execution.

  So we actually have two acquisition funnels, not one:

  ┌───────────────┬─────────────────────────────────────────────────────────────┬────────────────────────────────────┬────────────────────────────────────────────┬──────────┐
  │    Funnel     │                            User                             │                Pain                │                  Solution                  │ Priority │
  ├───────────────┼─────────────────────────────────────────────────────────────┼────────────────────────────────────┼────────────────────────────────────────────┼──────────┤
  │ A: "Blocked"  │ Technical user, had OpenClaw working on Claude, got blocked │ Model switching is too complex     │ CLI wizard: proclaw setup --model deepseek │ P0       │
  ├───────────────┼─────────────────────────────────────────────────────────────┼────────────────────────────────────┼────────────────────────────────────────────┼──────────┤
  │ B: "Too Hard" │ Non-technical user, wants OpenClaw but can't deploy it      │ Deployment requires tech expertise │ Dashboard: 2 clicks → running agent        │ P0       │
  └───────────────┴─────────────────────────────────────────────────────────────┴────────────────────────────────────┴────────────────────────────────────────────┴──────────┘

  Funnel B is potentially much bigger than Funnel A. Funnel A is dozens per day. Funnel B is potentially thousands — every person who starred OpenClaw but never got it running.

  And here's the kicker: Funnel B users are more valuable. They're less price-sensitive (they're not CLI experts who can DIY), they're stickier (they depend on the UI), and they upgrade faster (they need more features because they can't
  self-serve).

  Revised Sequencing

  Given this, here's my proposed priority order:

  Week 1: Ship the "Unblocking" MVP (Funnel A)

  ┌─────┬────────────────┬────────────────────────────────────────────────────────────────────────────┐
  │ Day │     Module     │                                 What Ships                                 │
  ├─────┼────────────────┼────────────────────────────────────────────────────────────────────────────┤
  │ 1-2 │ Model profiles │ Pre-tested configs for DeepSeek, Together, Groq, Ollama in model_gateway/  │
  ├─────┼────────────────┼────────────────────────────────────────────────────────────────────────────┤
  │ 2-3 │ CLI wizard     │ proclaw setup → interactive model selection → config → health check → done │
  ├─────┼────────────────┼────────────────────────────────────────────────────────────────────────────┤
  │ 3-4 │ Community tier │ $0 plan in billing, feature gates, no credit card                          │
  ├─────┼────────────────┼────────────────────────────────────────────────────────────────────────────┤
  │ 4-5 │ Landing page   │ Static "Blocked by Anthropic?" page, SEO content                           │
  └─────┴────────────────┴────────────────────────────────────────────────────────────────────────────┘

  Outcome: Technical users can pip install proclaw && proclaw setup and be running in 3 minutes. Start acquiring Funnel A immediately.

  Week 2: Ship the "Magic Dashboard" (Funnel B)

  ┌─────┬─────────────────────────────┬──────────────────────────────────────────────────────────────────┐
  │ Day │           Module            │                            What Ships                            │
  ├─────┼─────────────────────────────┼──────────────────────────────────────────────────────────────────┤
  │ 1-2 │ Dashboard onboarding wizard │ /onboarding — 3-step: pick model → name your agent → launch      │
  ├─────┼─────────────────────────────┼──────────────────────────────────────────────────────────────────┤
  │ 2-3 │ Model switcher page         │ /settings/model — one-click switch, health check                 │
  ├─────┼─────────────────────────────┼──────────────────────────────────────────────────────────────────┤
  │ 3-4 │ "Deploy in 2 clicks" flow   │ Hosted deployment — sign up → dashboard → agent running          │
  ├─────┼─────────────────────────────┼──────────────────────────────────────────────────────────────────┤
  │ 4-5 │ Polish + video              │ 3-min demo video, landing page update with dashboard screenshots │
  └─────┴─────────────────────────────┴──────────────────────────────────────────────────────────────────┘

  Outcome: Non-technical users can sign up, click twice, and have a running agent. This is the "magic" the Discord users are asking for.

  Week 3: Harden + Grow

  - Model compatibility benchmarks (publish data)
  - 30-day bundled model trial logic
  - Launch promo (200 calls/day for 90 days)
  - SEO content (blog posts, guides)

  On the Discord Goldmine for ClawEngine

  You're absolutely right — that Discord channel is the best PRD research you'll ever get. Real users, real problems, unprompted. The issues they're reporting (connectivity, captcha navigation, deployment complexity) are exactly the
  things ClawEngine should solve natively.

  I'd suggest we start a ClawEngine requirements capture as a separate doc — docs/clawengine_requirements.md — and systematically mine the Discord for:

  ┌─────────────────────┬─────────────────────────────────────────────────────────────┬───────────────────────────────────────────────────────────────────────────────┐
  │      Category       │                     Discord Pain Points                     │                            ClawEngine Opportunity                             │
  ├─────────────────────┼─────────────────────────────────────────────────────────────┼───────────────────────────────────────────────────────────────────────────────┤
  │ Deployment          │ "How do I install this?" "Docker not working" "WSL issues"  │ One-binary distribution, native Windows support                               │
  ├─────────────────────┼─────────────────────────────────────────────────────────────┼───────────────────────────────────────────────────────────────────────────────┤
  │ Captcha/Navigation  │ "Agent can't get past captcha" "Cloudflare blocks it"       │ Native browser integration with captcha-aware retry, headless Chrome profiles │
  ├─────────────────────┼─────────────────────────────────────────────────────────────┼───────────────────────────────────────────────────────────────────────────────┤
  │ Connectivity        │ "API keeps disconnecting" "WebSocket drops"                 │ Built-in connection resilience, automatic reconnect, offline queue            │
  ├─────────────────────┼─────────────────────────────────────────────────────────────┼───────────────────────────────────────────────────────────────────────────────┤
  │ Model compatibility │ "Tool calls fail on DeepSeek" "Ollama doesn't return JSON"  │ Native multi-model adapter layer, graceful degradation                        │
  ├─────────────────────┼─────────────────────────────────────────────────────────────┼───────────────────────────────────────────────────────────────────────────────┤
  │ Tool reliability    │ "Shell command hangs" "File read fails silently"            │ Timeout-aware tool execution, structured error propagation                    │
  ├─────────────────────┼─────────────────────────────────────────────────────────────┼───────────────────────────────────────────────────────────────────────────────┤
  │ Memory/Context      │ "Agent forgets what it was doing" "Context window overflow" │ Native conversation management with summarization                             │
  └─────────────────────┴─────────────────────────────────────────────────────────────┴───────────────────────────────────────────────────────────────────────────────┘

  Every one of these is a feature OpenClaw does poorly that ClawEngine can do right — because we're building from scratch with these problems in mind instead of bolting fixes onto a convenience-first architecture.

  Want me to create the clawengine_requirements.md doc now and capture this framework? Or should we table that and jump straight into building the 4 critical modules?

  My suggestion: I create a quick skeleton clawengine_requirements.md (5 minutes), then we go heads-down on the modules. The Discord mining can happen in parallel as a background task while we build.