Skip to main content
Glama
Sign In

MagicPixel

by art.magicpixel
OverviewDiscussionsAdmin

Generate pixel art

generate_pixel_art

Generate new pixel art from a text prompt (in-app Generate → New). Infer sensible defaults and state them instead of interrogating the user: size = the width/height of the project's existing assets (list_assets), else 64; density quality; subject_category inferred from the prompt (character/item/building/ui/texture); no style reference unless the user mentions one (reference_role style). Texture needs 256+. Returns up to four labelled previews and saves every result. ~8 credits (HD). For a ready-made asset, list_marketplace_assets searches the asset library.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "prompt": "<prompt>"
    },
    "name": "generate_pixel_art"
  }
}

Input Schema

NameRequiredDescriptionDefault
nameNoLibrary asset name when saving (default derived from prompt).
saveNoSave to the user's library (default true). When false, only returns the PNG.
sizeNoTarget square artboard size in pixels. Allowed: 64, 128, 256, 512, 1024. Match the project's existing assets (list_assets width/height); default 64. Every sprite size runs the same calibrated HD path (512/1024 are the 256px result enlarged losslessly). Texture subject needs 256+.
promptYesSubject only (e.g. 'red potion bottle'). Put canvas size in `size`, palette cap in `color_policy` — not in this text. Avoid '64x64' / 'no anti-aliasing' / 'crisp pixel clusters' lectures (pipeline owns those). Same cap as the in-app prompt box: longer text is trimmed to the first 1000 characters (60 for a texture material name) and the reply says so.
densityNo"quality" (default, about 3 results) at every size; "quantity" (as many as fit) only at 64. Generate has no "max": it and any density the size does not offer run as quality. Ignored for texture (always one tile).
folder_idNoLibrary folder id from create_folder (scope must be 'art'). Places the saved artboard in that folder and inherits the folder's project_id when project_id is omitted.
project_idNoProject to attach the saved asset to. Defaults to the folder's project (if folder_id), else the user's oldest project. Required for the asset to appear in MagicPixel's project-scoped library.
color_policyNoPalette policy matching in-app ColorPolicyControl: "32"|"64"|"128"|"256" (max colors) or "none" (don't reduce). Default "128". Put palette caps here, not in the prompt.
project_styleNoStyle for this generation: "project" (default) draws the result in the game world of the account's active style (its brief, see get_style_profile); "none" leaves the brief out for a one-off asset outside that world.
reference_roleNoWith style_reference_id: "style" (default, borrow look only) or "subject" (keep the reference subject).
style_image_idNoOptional: one image of the active style's kit to anchor on (an id from get_style_set, at this size). Its look, scale and padding carry over and its type replaces subject_category. Omit to use subject_category's first image.
subject_categoryNoSubject: character|item|building|ui|custom|texture. Infer from the prompt; default character. Texture requires size 256+.
style_reference_idNoOptional style reference (upload_style_reference / list_style_references). Omit unless the user mentions a reference.

Schema Changelog

Changes observed during successful MCP inspections.

No schema history has been recorded yet.

TDQS

A4.3/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare openWorldHint=true and destructiveHint=false, so the description is not carrying the full safety burden, but it adds substantive behavior anyway: it returns up to four labelled previews, saves every result, costs ~8 credits (HD), enforces 256+ for texture, and notes prompts are trimmed at 1000 chars. Missing only whether saving can be reverted or how failures are surfaced.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded with purpose, then defaults, then return/cost, then the alternative — a sensible ordering. The semicolon-chained default rules are dense and read as a run-on, costing a little clarity, but almost no sentence is wasted.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a 13-parameter generation tool with no output schema, the description covers cost, output shape (up to four labelled previews), persistence, and default-selection logic. The remaining gap is minor: it never says how to retrieve results afterwards (get_generation_result) or what a failed generation returns.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so 3 is the floor, and the description genuinely adds policy beyond it: the default-inference order (existing assets then 64), the rule that style references are only used when the user mentions one, and the cross-tool link to list_assets for sizing. It does not explain interactions between folder_id and project_id beyond what the schema already states.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb and resource ("Generate new pixel art from a text prompt") plus the in-app path (Generate → New), and distinguishes itself from the ready-made path by naming list_marketplace_assets. An agent can separate this from generate_variants or reskin_character without opening a schema.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Gives strong contextual guidance: infer and state defaults instead of interrogating the user, with concrete rules per parameter (size from list_assets, density quality, subject_category from the prompt, no style reference unless mentioned). It also names an alternative for the ready-made case, but never sets an explicit boundary against generate_variants or generate_directions.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All MagicPixel toolsInspect
Try in Browser
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources