{ "markdown": "
\n\n\n\n\"prompt-to-asset\n\n

prompt‑to‑asset

\n\n

Free image generation for developers, from the CLI or your AI assistant.
\nTurn one brief into app icons, favicons, OG images, logos, splash screens, SVG, and platform bundles using the best zero-cost routes first.
\nBuilt for Cloudflare Workers AI, NVIDIA NIM, HF Inference, Stable Horde, Pollinations, free trials, and offline SVG.

\n\n

\n \"npm\"\n \"downloads\"\n \"CI\"\n \"license\"\n \"node\"\n \"MCP\"\n \"Glama\n \"zero-key\"\n \"stars\"\n

\n\n

\n \"Install\n \"Install\n \"Claude\n \"Smithery\"\n

\n\n\n Quickstart ·\n Free paths ·\n Usage ·\n Modes ·\n Free APIs ·\n MCP tools ·\n Full guide ·\n Changelog\n\n\n
\n\n---\n\n## 30-second start\n\n**Pick one. Run it. You're done.** The recommended path is first.\n\nRuntime: **Node ≥ 20.11** (24 recommended). macOS, Linux, Windows (WSL2 for `--fix` native deps).\n\n\n\n\n\n\n\n
\n\n### 🔵 AI assistant · recommended\n\nClick an install button above. Then paste one of these into chat:\n\n```text\n• Make a transparent logo for Forge, a dev-tools brand. Flat vector, warm orange.\n• Make a favicon for my app, dark-mode aware.\n• Fan this master.png out to iOS + Android + PWA.\n• Ingest this screenshot as a mark, vectorize, export everything.\n```\n\nWorks in Cursor, Claude Code, VS Code, Windsurf, Codex, Gemini CLI. Zero terminal typing.\n\n\n\n### 🟢 Zero key, zero install\n\nOne `curl`, one `npx`. Offline fan-out to every platform.\n\n```bash\ncurl -o logo.png \\\n \"https://image.pollinations.ai/prompt/\\\nminimal+flat+vector+logo\\\n?model=flux&width=1024&nologo=true\"\n\nnpx prompt-to-asset export logo.png \\\n --platforms ios,android,pwa,favicon\n```\n\nOutputs: iOS AppIconSet, Android adaptive, PWA, favicon bundle, visionOS scaffold.\n\n\n\n### 🟣 CLI · free API keys\n\nBest CLI experience: free tokens first, paid keys optional.\n\n```bash\nnpm i -g prompt-to-asset\np2a doctor # check env\np2a doctor --fix # auto-install deps\np2a pick # interactive\n```\n\nStart with Cloudflare Workers AI, NVIDIA NIM, HF Inference, Stable Horde, and free trials. Paid providers remain optional fallbacks.\n\n
\n\n\n\n### Free paths at a glance\n\nYou don't need a paid API key. Ranked best-first:\n\n| Route | Gets you | Signup |\n| ------------------------------------ | -------------------------------------- | ----------------------- |\n| **`inline_svg`** via AI assistant | Logos, favicons, icon packs — instant | None |\n| **Cloudflare Workers AI** | Flux-1-Schnell + SDXL, 10k neurons/day | Free token + account ID |\n| **NVIDIA NIM** | Flux.1-dev, Flux.2-klein, SDXL, SANA | Free token, no card |\n| **HF Inference** | SDXL, SD3, Flux dev/schnell | Free read token |\n| **Stable Horde** | SDXL, Flux on community GPUs | Anonymous queue |\n| **Pollinations** (HTTP GET) | Last-resort raster, RGB | None |\n| **Google AI Studio** (paste-only UI) | Nano Banana / Nano Banana Pro | Google account |\n\nDetails + quotas: [Free paths beyond Pollinations](#free-paths-beyond-pollinations). Run `p2a doctor` or ask your assistant for `asset_doctor()` to see what's live right now.\n\n> [!TIP]\n> Stuck? Click **Install in Cursor** or **Install in VS Code** above, restart the editor, and say: _\"make a favicon for my app, dark-mode aware.\"_\n\n---\n\n## Highlights\n\n- **Free CLI-first generation** — `p2a doctor`, `p2a pick`, and the MCP tools rank free API routes before paid models, so developers can generate inside the CLI without enabling billing.\n- **[Three execution modes](#the-three-modes)** — `inline_svg` (host LLM authors SVG), `external_prompt_only` (paste into any web UI), `api` (server calls a free or paid provider). Pick what fits. All three can finish on $0.\n- **[60+ models, free-first routing](#the-router)** — Cloudflare Workers AI, NVIDIA NIM, HF, Horde, Pollinations, free trials, plus optional paid OpenAI / Ideogram / Recraft / BFL / Gemini. Each rule cites a research source.\n- **[Refuses to do the wrong thing](#the-router)** — the `Never` column. No wordmarks past 3 words through a diffusion sampler. No transparent PNG through Imagen. No `negative_prompt` on Flux.\n- **[Offline platform fan-out](#platform-support)** — one 1024² master → iOS AppIconSet, Android adaptive + monochrome, PWA 192/512/512-maskable, favicon bundle, visionOS parallax, Flutter launcher. Zero network.\n- **[Validates before shipping](#security)** — tier-0 checks on every output: dims, alpha presence, checkerboard FFT, safe-zone bbox, ΔE2000 palette drift, WCAG contrast, OCR Levenshtein on wordmarks.\n- **[Sprite sheets + 9-slice](#mcp-tools)** — pack PNG frames into TexturePacker-compatible atlases (Phaser, PixiJS, Godot, Unity); emit 9-slice numbers + CSS `border-image` + Android `.9.png`.\n\n---\n\n## Table of contents\n\n
\nClick to expand\n\n- [30-second start](#30-second-start)\n- [Usage](#usage) — what to say to your assistant\n- [The three modes](#the-three-modes) — inline_svg · external_prompt_only · api\n- [The router](#the-router) — which model for which job, and what never\n- [Free paths beyond Pollinations](#free-paths-beyond-pollinations)\n- [Install](#install)\n- [Models covered](#models-covered) — 60+\n- [MCP tools](#mcp-tools) — 25 tools\n- [CLI surface](#cli-surface)\n- [Brand bundle (`brand.json`)](#brand-bundle)\n- [Platform support](#platform-support)\n- [Architecture](#architecture)\n- [Comparison](#comparison)\n- [Security](#security)\n- [Research-backed decisions](#research-backed-decisions)\n- [Development](#development)\n- [Community](#community) · [License](#license)\n\n
\n\n---\n\n## Usage\n\nYou just talk to your assistant. Example from a new chat:\n\n> **Make me a transparent logo for a developer-tools company called Forge. Flat vector, two-tone warm orange on neutral.**\n\nBehind the scenes:\n\n1. **`asset_doctor()`** — check what modes and providers are live.\n2. **`asset_init_brand({ app_name: \"Forge\", palette: [\"#EA580C\", \"#F5F5F4\"] })`** if no `brand.json` exists.\n3. **`asset_enhance_prompt({ brief })`** returns an `AssetSpec`: classification, rewritten prompt, `modes_available[]`, optional `svg_brief`, optional `paste_targets`, and a `routing_trace` pointing at the research file that backed the decision (plus `never_models` — why Imagen or DALL·E got rejected).\n4. Assistant offers you **`inline_svg`** / **`external_prompt_only`** / **`api`**.\n5. If **`inline_svg`**: it writes `` inline and calls `asset_save_inline_svg` → writes `master.svg`, `favicon.ico`, apple-touch, AppIconSet, PWA bundle to disk.\n6. If **`external_prompt_only`**: assistant shows the refined prompt and the best paste target (free first). You generate, save, then say _\"ingest this file\"_ → `asset_ingest_external`.\n7. If **`api`**: assistant calls the routed provider. Server mattes, vectorizes, exports, validates.\n8. Follow-up: _\"also fan this out for iOS and Android\"_ → `asset_export_bundle` with the saved master.\n\nZero CLI typing. The CLI is still first-class for CI, shell scripts, and non-MCP environments — both surfaces hit the same core.\n\n---\n\n## Why this exists\n\nTwo facts shape everything here.\n\n> **Producing production-grade software assets is a routing and post-processing problem, not a prompt-engineering problem.**\n\nImagen 3/4 and Gemini Flash Image can't produce real RGBA PNGs — their VAE is RGB-only, so asking for a transparent background renders the grey-and-white checkerboard _as pixels_. SDXL can't spell past ~8 characters. Only Recraft emits native SVG. Flux errors on `negative_prompt`. None of that is visible in the model UI. All of it silently breaks one-shot \"prompt → asset\" tools.\n\n> **You may not have an image-model API key. The plugin works anyway.**\n\nEvery one of the three modes can finish on $0.\n\n---\n\n## The three modes\n\n```mermaid\nflowchart LR\n A[\"one-line brief\"] --> B[\"asset_enhance_prompt\"]\n B --> C{\"mode\"}\n C -->|inline_svg| D[\"Host LLM emits SVG inline
→ asset_save_inline_svg\"]\n C -->|external_prompt_only| E[\"Paste into web UI
→ asset_ingest_external\"]\n C -->|api| F[\"Server calls routed provider\"]\n D --> G[\"matte · vectorize · validate\"]\n E --> G\n F --> G\n G --> H[\"AssetBundle
ios · android · pwa · favicon · visionos · flutter\"]\n```\n\n| Mode | Key? | What happens | Best for |\n| -------------------------- | -------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | -------------------------------------------------------------- |\n| **`inline_svg`** | No | Server returns an SVG-authoring brief (viewBox, palette, path budget ≤ 40). Host LLM emits `…` inline, then `asset_save_inline_svg` writes master + favicon.ico + apple-touch + AppIconSet + PWA bundle. Instant. Deterministic. | Logos, favicons, icon packs, stickers, simple app-icon masters |\n| **`external_prompt_only`** | No | Server returns the dialect-correct prompt plus a ranked list of paste targets, free paths first: Pollinations, HF Inference, Stable Horde, Google AI Studio, Ideogram, Recraft, Midjourney, fal.ai, BFL, ChatGPT, Firefly, Krea. Generate elsewhere, save locally, call `asset_ingest_external`. | Anything — best for illustrations, heroes, text-heavy logos |\n| **`api`** | Optional | Server calls the provider directly. Works **zero-key via Pollinations / Horde / HF**, or with paid keys. Route → generate → matte → vectorize → export → validate → content-addressed bundle. | Automation, CI, no rate-limit tolerance |\n\nThe host LLM picks the mode, or you do. The server surfaces `modes_available` so the assistant offers them to you. **Free paths first — always.**\n\n---\n\n## The router\n\nRouter decisions live in [`data/routing-table.json`](./data/routing-table.json). Capability matrix in [`data/model-registry.json`](./data/model-registry.json). **Every rule cites its research source.**\n\n| Need | Primary | Fallback | Never |\n| ------------------------ | --------------------------------------------------------------- | ------------------------------------------------------- | ---------------------------------- |\n| Transparent PNG mark | `gpt-image-1.5` with `background:\"transparent\"` | Ideogram 3 Turbo (`/generate-transparent`) → Recraft V4 | Imagen, Gemini Flash Image, SD 1.5 |\n| Logo with 1–3 word text | Ideogram 3 Turbo → `gpt-image-1.5` → Recraft V4 | Composite SVG type over mark | Imagen, SD 1.5, `flux-schnell` |\n| Logo with >3 word text | **Never a diffusion sampler.** Mark + SVG typography composite. | — | — |\n| Native SVG | Recraft V4 (V3 for brand-style pipelines) | `inline_svg` (host LLM authors SVG) | Everyone else |\n| Photoreal hero | Free CLI route first: Cloudflare / NVIDIA NIM / HF | Flux Pro / `gpt-image-1.5` / Gemini with paid keys | DALL·E 3, Imagen 4 for new work |\n| Iterate an existing mark | `flux-kontext-pro` (edit-only) | Pollinations Kontext (free) | — |\n| Zero-cost everything | `inline_svg` → Cloudflare / NVIDIA NIM / HF / Horde | Pollinations → paste-only | — |\n\nThe **Never** column matters. It's why `prompt-to-asset` refuses to render wordmarks past 3 words in any diffusion sampler, and why asking for a transparent PNG never goes to Imagen.\n\n---\n\n## Free paths beyond Pollinations\n\n| Option | How | Best at | Catch |\n| -------------------------- | ---------------------------------------------- | ------------------------------------ | -------------------------------------------------------------------- |\n| **Cloudflare Workers AI** | Free API token + account ID | Flux-1-Schnell, SDXL, DreamShaper | 10k neurons/day cap (~900 Flux-Schnell or 5k SDXL-Lightning) |\n| **NVIDIA NIM** | Free API token, no card | Flux.1-dev, Flux.2-klein, SDXL, SANA | 1k requests/month; strong free CLI route |\n| **HF Inference** | Free read token | SDXL, SD3, Flux dev + schnell | Rate-limited, cold-start latency |\n| **Stable Horde** | Anonymous kudos queue | SDXL, Flux community GPUs | Minutes of queue on the free lane |\n| **Pollinations.ai** | `curl` → HTTP GET. No signup. | Throwaway raster | Last resort: silent model swaps, downsizing, RGB only |\n| **Google AI Studio (UI)** | Free interactive web UI at aistudio.google.com | Nano Banana / Nano Banana Pro | No free API — paste-only; download PNG, call `asset_ingest_external` |\n| **Local ComfyUI** | Community `comfyui-mcp` adapter | Full fidelity, no caps | You bring the GPU |\n| **`inline_svg`** | Host LLM emits `` in chat | Logos, favicons, simple icons | ≤40 paths; simple geometry |\n| **`external_prompt_only`** | Paste into any web UI | Whatever that UI gives you | Manual save, then `asset_ingest_external` |\n\n
\n
Google Gemini / Imagen — verified pricing (Apr 2026)\n\nVerified 2026-04-26 against public Gemini API pricing:\n\n- **Imagen 4 image output is paid API.** The pricing table lists Free Tier as \"Not available\" and prices Fast / Standard / Ultra at $0.02 / $0.04 / $0.06 per image.\n- **Nano Banana family is paid API.** `gemini-3.1-flash-image-preview` (Nano Banana 2), `gemini-3-pro-image-preview` (Nano Banana Pro), and `gemini-2.5-flash-image` (original Nano Banana) list image output only in the paid tier. Billing must be enabled on the GCP project to use these via API.\n- **Free for text, multimodal understanding, and embeddings.** The Gemini text-out models still have `Free of charge` input + output on the free tier.\n- **Free interactive image generation is also via the AI Studio web UI** at [aistudio.google.com](https://aistudio.google.com). Community-observed limit 500–1,000 images/day, dynamic. Use `external_prompt_only` + `asset_ingest_external` if you want Nano Banana quality without billing.\n- **Free image generation via the Gemini consumer app** at [gemini.google.com](https://gemini.google.com): Basic 20/day, AI Plus 50/day, AI Pro 100/day, Ultra 1,000/day.\n- **Paid API pricing (per image, standard):** Nano Banana (`gemini-2.5-flash-image`) $0.039; Nano Banana 2 Flash (`gemini-3.1-flash-image-preview`) $0.045/0.5K, $0.067/1K, $0.101/2K, $0.151/4K; Nano Banana Pro (`gemini-3-pro-image-preview`) $0.134/1K-2K, $0.24/4K (+ $0.0011 per input image); Imagen 4 Fast $0.02, Standard $0.04, Ultra $0.06. Batch API is 50% off.\n\n\n\n> Run `p2a doctor` (or ask your assistant for `asset_doctor()`) to see what's live in your environment right now.\n\n---\n\n## Install\n\nEvery command works via `npx` — no install required.\n\n```bash\n# Zero install\nnpx prompt-to-asset doctor # what's live in this shell right now\nnpx prompt-to-asset doctor --fix # auto-install native deps (brew / cargo / scoop; never sudo)\nnpx prompt-to-asset pick # interactive route picker\nnpx prompt-to-asset init --register # scaffold brand.json + register in .cursor / .vscode / .windsurf\n\n# Or global for daily use\nnpm i -g prompt-to-asset\np2a doctor\n\n# Or per-project for CI\nnpm i -D prompt-to-asset\n```\n\n### Register with your AI assistant\n\n\n\n\n\n\n
\n\n**Claude Code**\n\n```bash\nclaude mcp add prompt-to-asset -- p2a\n```\n\n**Smithery (universal)**\n\n```bash\nnpx -y @smithery/cli install prompt-to-asset --client claude\n```\n\n\n\n**Cursor · VS Code · Windsurf · Codex · Gemini CLI**\n\nUse the **install buttons at the top** of this README, or see [`docs/install.md`](./docs/install.md) for the exact stanza per IDE.\n\n**Claude Desktop**\n\nDownload the [`.mcpb` bundle](https://github.com/MohamedAbdallah-14/prompt-to-asset/releases/latest) → double-click → restart.\n\n
\n\nOnce registered, your assistant has the full **24 `asset_*` tool** surface.\n\n---\n\n## Models covered\n\n**Paid direct APIs:** `gpt-image-1`, `gpt-image-1.5`, `dall-e-3` (deprecated 2026-05-12), `imagen-3`, `imagen-4`, `gemini-3-flash-image` (Nano Banana), `gemini-3-pro-image`, `sd-1.5`, `sdxl`, `sd3-large`, `playground-v3`, `flux-schnell`, `flux-dev`, `flux-pro`, `flux-2`, `flux-kontext-pro`, `ideogram-3`, `ideogram-3-turbo`, `recraft-v3`, `leonardo-phoenix`, `leonardo-diffusion-xl`, `fal-flux-pro`, `fal-flux-2`, `fal-sdxl`.\n\n**Free-tier / zero-key:** `pollinations-flux`, `pollinations-turbo`, `pollinations-kontext`, `pollinations-sd`, `horde-sdxl`, `horde-flux`, `hf-sdxl`, `hf-sd3`, `hf-flux-schnell`, `hf-flux-dev`.\n\n**Paste-only surfaces:** `midjourney-v6`, `midjourney-v7`, `firefly-3`, `krea-image-1`. Calling `asset_generate_*` with `mode: \"api\"` against a paste-only primary soft-falls-back to the first API-reachable model in the chain and surfaces a warning. If the whole chain is paste-only, you get an `ExternalPromptPlan` rather than an error.\n\n---\n\n\n\n
\n🛠 MCP tool surface (25 tools)\n\n| Tool | Purpose |\n| ------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| `asset_capabilities` | Inventory of modes + providers. Buckets paid / free-tier / paste-only; surfaces zero-key routes first. Read-only. |\n| `asset_enhance_prompt` | Classify, route, rewrite. Returns modes + `svg_brief` + `paste_targets` + `routing_trace { research_sources, never_models, fallback_chain }` + `clarifying_questions[]` when the brief is ambiguous. Read-only. |\n| `asset_generate_logo` | `inline_svg` / `external_prompt_only` / `api`. Returns `InlineSvgPlan` / `ExternalPromptPlan` / `AssetBundle`. |\n| `asset_generate_app_icon` | Same three modes. `api` produces full iOS / Android / PWA / visionOS / Flutter fan-out. Set `ios_18_appearances: true` for dark + tinted variants. |\n| `asset_generate_favicon` | `favicon-{16,32,48}.png`, `icon.svg`, `icon-dark.svg`, `apple-touch`, PWA 192/512/512-maskable, `` snippet, `manifest.webmanifest`. |\n| `asset_generate_og_image` | 1200×630 via Satori + `@resvg/resvg-js`. Deterministic typography, no diffusion-rendered text garbage. |\n| `asset_generate_illustration` | `external_prompt_only` / `api`. Brand-locked via bundle refs, LoRA, or `style_id`. Routed primary: Flux.2 (up to 8 refs). |\n| `asset_generate_splash_screen` | iOS `LaunchScreen-2732.png`, Android `mipmap-*/splash.png` + theme XML, PWA splash + README. Pass `existing_mark_svg` to reuse an approved mark. |\n| `asset_generate_hero` | Marketing hero art (16:9 / 21:9 / 3:2 / 2:1). `external_prompt_only` / `api`. |\n| `asset_save_inline_svg` | Round-trip for `inline_svg`. Validates the SVG against the brief, writes the bundle. |\n| `asset_ingest_external` | Round-trip for `external_prompt_only`. Matte → vectorize → validate → bundle. |\n| `asset_remove_background` | BiRefNet / BRIA RMBG-2.0 / LayerDiffuse / difference matte / U²-Net. |\n| `asset_vectorize` | `vtracer` / `potrace` / Recraft / posterize fallback, then SVGO. |\n| `asset_upscale_refine` | DAT2 / Real-ESRGAN / SUPIR / img2img / Lanczos; asset-type-aware. |\n| `asset_validate` | Tier-0 (dims, alpha, checkerboard FFT, safe-zone bbox, ΔE2000 palette, WCAG contrast, OCR Levenshtein). Tier-2 VLM-as-judge via `PROMPT_TO_BUNDLE_VLM_URL`. |\n| `asset_brand_bundle_parse` | Parse `brand.json` / DTCG tokens / AdCP / Markdown into a canonical `BrandBundle`. |\n| `asset_doctor` | Structured env inventory: native deps, free-tier routes ranked best-first, paid keys, paste-only surfaces, pipeline URLs, mode flags, \"what to try next.\" Read-only. |\n| `asset_models_list` | Browse the 60+ model registry with filters: `free` / `paid` / `paste_only` / `rgba` / `svg`. Read-only. |\n| `asset_models_inspect` | Full capability dump for one model id (or aka alias). Strengths, weaknesses, paste targets, routing rules, env status. Read-only. |\n| `asset_export_bundle` | Fan a 1024² master PNG into iOS AppIconSet + Android adaptive + PWA maskable + visionOS parallax + Flutter launcher + favicon. Offline. |\n| `asset_sprite_sheet` | Pack PNG/WEBP/JPG frames into a sprite sheet + TexturePacker-compatible JSON atlas (Phaser / PixiJS / Godot / Unity). Offline. |\n| `asset_nine_slice` | Emit a 9-slice config + CSS `border-image` + engine-ready numbers (Unity / Godot / Phaser / PixiJS) from an image plus four pixel guides. Optional Android `.9.png`. |\n| `asset_init_brand` | Scaffold `brand.json` and ensure the assets dir exists. Auto-detects Next.js, Expo, Flutter, Xcode, Astro, Vite, Remix, Nuxt, React Native, Electron, Node. |\n| `asset_train_brand_lora` | Wrap a user-owned LoRA training endpoint (`PROMPT_TO_BUNDLE_MODAL_LORA_TRAIN_URL`). Path-guarded; validates inputs. |\n\nTools are annotated `readOnlyHint` / `idempotentHint` so Cursor auto-approves without prompting.\n\n
\n\n\n\n
\n⌨️ CLI surface\n\nUsed by the LLM over `Bash` when MCP isn't registered yet, and by CI. Every read-only command accepts `--json`.\n\n```\np2a # default — MCP stdio server\np2a mcp # same, explicit\np2a export # offline platform fan-out\np2a export --json\np2a init # interactive brand.json + IDE registration hints\np2a init --register # + auto-write .cursor/mcp.json / .vscode/mcp.json / .windsurf/mcp.json\np2a pick # interactive model picker\np2a doctor # environment inventory\np2a doctor --json # structured output\np2a doctor --data # check data/model-registry.json ↔ data/routing-table.json consistency\np2a doctor --fix # auto-install missing native deps (brew / cargo / scoop; never sudo)\np2a models list # --free | --paid | --paste-only | --rgba | --svg\np2a models inspect # full capability dump\np2a sprite-sheet # pack frames → PNG + atlas\np2a nine-slice # 9-slice JSON + CSS + engine numbers + .9.png\np2a --help\n```\n\n
\n\n\n\n
\n🎨 Brand bundle — brand.json\n\n```json\n{\n \"name\": \"Halcyon\",\n \"palette\": [\"#2563eb\", \"#ffffff\"],\n \"fonts\": { \"display\": { \"family\": \"Inter\", \"weights\": [700, 800] } },\n \"style_refs\": [\"https://…/sample1.png\", \"./refs/style2.png\"],\n \"do_not\": [\"drop shadows\", \"heavy gradients\"],\n \"lora\": \"halcyon-flux-v2\",\n \"sref_code\": \"--sref 1234567890\",\n \"style_id\": \"rc_halcyon_01\"\n}\n```\n\n`p2a init` writes this for you, detecting the framework and suggesting an assets directory. Once present, every generator reads from it automatically.\n\n
\n\n\n\n
\n📱 Platform support\n\n| Platform | What you get |\n| -------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| **iOS (Xcode)** | `AppIcon.appiconset` with 1024 opaque, squircle-ready. iOS 18 dark + tinted variants via `ios_18_appearances: true`. |\n| **Android** | Adaptive foreground + background, Android 13 monochrome, all mipmap densities, optional `.9.png`. |\n| **PWA / web** | `favicon.ico` (16/32/48 multi-res), `icon.svg` with `prefers-color-scheme` dark support, `apple-touch-icon.png` 180×180 opaque, 192/512/512-maskable, `manifest.webmanifest`, `` snippet for your ``. |\n| **Flutter** | Pre-populated `flutter_launcher_icons.yaml` wiring iOS, Android adaptive, web, macOS, Windows. |\n| **visionOS** | Three-layer parallax scaffold with a README. Layer split stays a human decision. |\n| **Next.js / Astro / Vite / Remix / Nuxt / Expo / React Native / Electron** | Framework detection via `p2a init` / `asset_init_brand` and a sensible output dir. |\n| **Games** | `sprite-sheet` produces TexturePacker-compatible atlases (Phaser, PixiJS, Godot, Unity). `nine-slice` emits engine-ready numbers. |\n\n
\n\n\n\n
\n🏗 Architecture\n\n```\n brief (text)\n ↓ asset_capabilities → modes available + free/paid/paste-only bucketing\n ↓ asset_enhance_prompt → AssetSpec {\n ↓ routing_trace: { rule_id, reason, research_sources, never_models, fallback_chain },\n ↓ modes_available,\n ↓ svg_brief?, (inline_svg)\n ↓ paste_targets?, (external_prompt_only)\n ↓ rewritten_prompt, …\n ↓ }\n ↓\n ├─ mode: inline_svg → host LLM emits ; asset_save_inline_svg writes bundle\n ├─ mode: external_prompt_only → user pastes into web UI; asset_ingest_external runs matte → vectorize → validate\n └─ mode: api → provider(model, prompt, params) → matte → vectorize → upscale → export → validate\n```\n\nContent-addressed storage: `assets///.`. The MCP server is synchronous and stateless. `prompt_hash` and `params_hash` in every `AssetBundle` are designed to drop straight into a BullMQ / SQS / Cloudflare Queues `jobId` for a hosted pipeline. Reference design: [`docs/research/18-asset-pipeline-tools/18e-production-pipeline-architecture.md`](./docs/research/18-asset-pipeline-tools/18e-production-pipeline-architecture.md).\n\n### Design thesis\n\nYou own the API keys. The LLM owns everything else.\n\nThe only thing that happens in a terminal is installing the package and putting keys in `.env`. Secrets shouldn't pass through chat. Every other verb — doctor checks, model inspection, platform fan-out, brand scaffolding, sprite sheets, 9-slice configs — is an MCP tool the assistant calls when you ask in natural language.\n\n
\n\n---\n\n## Comparison\n\n| Tool | Prompt enhancement | Multi-model routing | Zero-key | Dev-asset bundle | Offline platform fan-out |\n| ---------------------------------------- | :----------------: | :-----------------: | :---------------------------------: | :--------------: | :----------------------------------------------------: |\n| Promptati / PromptHero | cinematic only | ✗ | ✗ | ✗ | ✗ |\n| Looka / Brandmark / Designs.ai | ✗ | ✗ | ✗ | partial | ✗ |\n| ChatGPT / Midjourney / Ideogram (direct) | ✗ | ✗ | ✗ | ✗ | ✗ |\n| appicon.co | ✗ | ✗ | ✓ | partial | iOS only |\n| flutter_launcher_icons | ✗ | ✗ | ✓ | partial | iOS + Android |\n| **`prompt-to-asset`** | ✓ | ✓ (60+ models) | ✓ (Pollinations / HF / Horde / SVG) | ✓ | ✓ (iOS + Android + PWA + visionOS + favicon + Flutter) |\n\n---\n\n## Security\n\nThis tool handles API keys for up to 15 providers. Non-negotiables:\n\n- **Keys live in env vars only.** Never written to disk, never logged, never echoed in MCP responses. Provider error bodies go through `redact()` (`packages/mcp-server/src/security/redact.ts`) before reaching the host LLM.\n- **Path access is allow-listed.** `image_path` / `output_dir` / `existing_mark_svg` resolve through symlinks and reject anything escaping project cwd + configured output dir + cache dir + OS tempdir. Widen with `P2A_ALLOWED_PATHS=/path1:/path2`.\n- **SVG is XSS-sanitized before any write.** `