# Cloudflare Free Tier Guard > Cost-aware Cloudflare implementation and review guidance for Workers, D1, R2, KV, Pages, Wrangler, and edge-hosted APIs. Use when Codex designs, edits, reviews, debugs, or deploys Cloudflare-related code or config, especially `wrangler.toml`, Worker entrypoints, storage bindings, upload flows, caching, or infrastructure choices where staying inside free-tier or low-cost limits matters. Use when this capability is needed. - Skill: `tomevault-io/cloudflare-free-tier-guard` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/cloudflare-free-tier-guard` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/cloudflare-free-tier-guard/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/cloudflare-free-tier-guard --- # Cloudflare Free Tier Guard ## Overview Apply a free-tier-first design review before making Cloudflare changes. Favor architectures that keep Workers thin, avoid unnecessary paid products, and make request, storage, and egress costs explicit. ## Quick Start 1. Read [references/free-tier-checklist.md](./references/free-tier-checklist.md). 2. If working in this repository's `server/` directory, read [references/launcherg-server-stack.md](./references/launcherg-server-stack.md). 3. Identify which Cloudflare primitives are involved: Workers, D1, R2, KV, Assets, Queues, Durable Objects, Cron, or custom domains. 4. State the expected hot path in one sentence: request source, Worker logic, storage access, response path. 5. Prefer the cheapest primitive that satisfies the requirement, then explain why. ## Workflow ### 1. Inventory the cost surface - Count likely request paths. - Separate metadata reads/writes from binary transfer. - Note whether traffic is public, authenticated, bursty, or background. - Call out which operations hit D1, R2, or Worker CPU. ### 2. Choose the lowest-cost architecture - Prefer: static assets via Assets/Pages, metadata in D1, blobs in R2, direct upload/download where possible. - Avoid adding Durable Objects, Queues, Cron, or extra Workers unless the requirement needs coordination, background execution, or fan-out. - Keep auth stateless when possible. Signed cookies or signed tokens are cheaper than session tables. - Keep the Worker off the binary data path unless authorization or transformation is required. ### 3. Apply free-tier techniques - Deduplicate uploads before generating storage writes. - Batch D1 writes and chunk large lookups. - Return only required columns. - Put TTLs on signed URLs and session cookies. - Cache safe reads aggressively. - Move heavy transforms to the client or build step. - Avoid polling loops and worker-to-worker hops. ### 4. Explicitly review risk before coding Before finalizing a design or patch, answer these questions: - Does this increase request count on every user action? - Does this push binary payloads through the Worker unnecessarily? - Does this add a paid Cloudflare product without proving it is needed? - Can the same behavior be achieved with Assets, D1, R2, cache headers, or client-side work? - What is the likely free-tier bottleneck first: request count, CPU, D1 operations, R2 storage, or egress? If the change is likely to exceed free-tier comfort, say so plainly and propose a cheaper alternative. ## Output Requirements When this skill is used, include a short cost note in the answer: - Architecture choice - Main free-tier pressure points - Concrete mitigations applied - Any remaining unknowns that need measurement ## References - [references/free-tier-checklist.md](./references/free-tier-checklist.md) - [references/launcherg-server-stack.md](./references/launcherg-server-stack.md) --- > Source: [ryoha000/launcherg](https://github.com/ryoha000/launcherg) — distributed by [TomeVault](https://tomevault.io).