Pricing V5: AI feature gating & cloud account tracking in CLI

[simple-agent-manager]

Context

Part of the Pricing V5 implementation. Source: strategy/pricing-implementation-checklist.md §4.4, §4.7, §4.8.

AI Feature Gating (§4.4)

Files: src/pkg/cli/generate.go:23-34, src/cmd/cli/command/debug.go:1-64

• Add client-side tier check for AI commands:
  • Starter: allow limited compose generation (show remaining quota), block or limit debugging
  • If at limit: "You've used your compose generations this month. Upgrade to Pro ($49/mo) for unlimited AI."
  • If debugging blocked: "AI debugging requires a Pro subscription ($49/mo)"
• Update defang generate help text to mention limits (Starter: limited, Pro: unlimited)
• Update defang debug help text to mention Pro requirement
• Note: defang generate is being deprecated in favor of compose-only generation (separate effort)

CanIUse Client (§4.7)

File: src/pkg/cli/client/caniuse.go:12-55

• Handle new CanIUseResponse fields (if added): allowed_modes, portal_access_level, ai_enabled, max_cloud_accounts/current_cloud_accounts

Cloud Account Tracking (§4.8)

CLI does NOT block deployments based on cloud account registration. Cloud accounts auto-register in Portal after deployment.

• Ensure provider_account_id is sent to Fabric on deployment (PutStack in stacks.go:30-53 doesn't include it — must add AccountInfo.AccountID)
• Show informational warning when deploying to a new account: "This deployment uses cloud account X. You currently have N accounts (limit: M). Additional accounts are billed at $29/mo."
• Add ErrCloudAccountLimit handling if Fabric enforces hard limits

Error Messages & Cost Estimation (§4.3, §4.6)

Files: src/pkg/cli/client/byoc/baseclient.go:41-85, src/pkg/cli/estimate.go:97-116

• Update GPU/scaling error messages: "GPU access requires a Pro subscription ($49/mo)"
• Update mode descriptions with pricing context: AFFORDABLE "(Available on all tiers)", BALANCED "(Requires Pro — $49/mo)", HA "(Requires Enterprise — $499/mo)"

🤖 Generated with Claude Code

[simple-agent-manager]

Decisions resolved (source)

• AI token limits on Starter: ~$1/mo equivalent (~5 complex compose file generations per month)