PixelLab Forge MCP
PixelLab Forge MCP is a server that lets AI assistants generate, edit, and manage pixel art assets (sprites, characters, animations, tilesets, UI, fonts) via the PixelLab API, with automatic saving and job tracking.
Image generation: Create pixel art from text descriptions, style references, or concept images using multiple engines (generate_image, generate_with_style, Pixen, Pixflux, Bitforge, Pro Flash).
Character & object creation: Build persistent characters/objects with 1/4/8 directional views, save them by ID, edit states, and tag or delete them.
Animation: Animate saved characters or objects via templates or custom text, generate keyframe interpolations, lip-sync/talking GIFs, and skeleton-based animations.
Tilesets & map assets: Create top-down, sidescroller, isometric, hex/octagon/square tiles, and map objects.
Editing & inpainting: Edit single or batch images, remove backgrounds, rotate views, resize, convert photos to pixel art, and perform mask-based region edits.
UI assets & fonts: Generate persistent UI panels, styled pixel-art fonts (glyph atlas + TTF), and convert between portraits and full-body sprites.
Prompt enhancement: Expand short descriptions into richer prompts for characters, animations, or Pixen images (text only).
Job management: Poll job status, list pending/history jobs, recover after disconnection, and retrieve saved assets.
File handling: Automatically saves PNGs to
./pixellab-forge-output/, supports reading saved images, and downloading character/object ZIPs or spritesheets.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@PixelLab Forge MCP Create a 32px tall pixel art knight with a shield."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
PixelLab Forge MCP
An MCP server that connects AI assistants to the PixelLab pixel art generation API. Generate sprites, tilesets, characters, animations, and more directly from Claude, Cursor, or any MCP-compatible client.
Generated images are automatically saved to ./pixellab-forge-output/ in your project directory, ready to be moved into your game assets.
Prerequisites
Node.js 18 or later
A PixelLab API key — get one at pixellab.ai/account
Related MCP server: aseprite-mcp
Setup
No installation needed. npx downloads and runs the package automatically on first use.
Package registry:
pixellab-forge-mcpis published to the public npm registry (npmjs.org) — that's whatnpx/npm installuse. It is not distributed via GitHub Packages, so the repo's/pkgs/npm/…page will 404; that's expected.
Claude Code (CLI)
claude mcp add pixellab-forge-mcp -e PIXELLAB_API_KEY=your-api-key-here -- npx pixellab-forge-mcpThis adds it to the current project. To make it available across all your projects:
claude mcp add pixellab-forge-mcp -s user -e PIXELLAB_API_KEY=your-api-key-here -- npx pixellab-forge-mcpThat's it. Claude Code will start the server automatically when you begin a conversation.
Claude Desktop
Add to your config file:
macOS:
~/Library/Application Support/Claude/claude_desktop_config.jsonWindows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"pixellab-forge-mcp": {
"command": "npx",
"args": ["pixellab-forge-mcp"],
"env": {
"PIXELLAB_API_KEY": "your-api-key-here"
}
}
}
}Cursor
Add to .cursor/mcp.json:
{
"mcpServers": {
"pixellab-forge-mcp": {
"command": "npx",
"args": ["pixellab-forge-mcp"],
"env": {
"PIXELLAB_API_KEY": "your-api-key-here"
}
}
}
}Other MCP Clients
Any MCP client that supports stdio transport can use PixelLab Forge. Set the command to npx pixellab-forge-mcp and pass PIXELLAB_API_KEY as an environment variable.
Generated Images
When a tool returns image data, PixelLab Forge automatically saves the images as PNGs to ./pixellab-forge-output/ in whatever directory the MCP server is running from (usually your project root).
Image inputs can be passed by file_path instead of inline base64 — any image argument accepts { "file_path": "pixellab-forge-output/sprite.png" }, which the server resolves from disk before calling the API (saves passing large base64 blobs through the model). Paths are restricted to the output directory and workspace root. The dedicated read_image tool remains available for loading an image explicitly.
Add this to your .gitignore:
pixellab-forge-output/Available Tools (103)
Generation tools automatically poll for results — no manual job status checking needed. If a job takes longer than 10 minutes or the connection drops, use list_pending_jobs to find the job ID and get_job_status to retrieve the result.
Image Generation
Tool | Description | Key Options |
| Generate pixel art from text |
|
| Match style from 1-4 references |
|
| Game UI elements (buttons, panels, icons) |
|
| Pixen engine (max area 512×512, dims ÷4) |
|
| Pixflux engine (32-400px) |
|
| Pixflux engine tuned for backgrounds/scenes | Same as |
| Bitforge engine (max 200px) |
|
Pro Flash
Newest engine. Native sizes 16×16 to 96×96 (custom 16–256, multiples of 4). Image results carry a durable source_image_id you can reuse to build a character or object from the same image without paying for it again.
Tool | Description | Key Options |
| Generate one native pixel-art image |
|
| Edit by text or reference; canvas never resizes |
|
| Repaint only white mask pixels |
|
| Saved character: first frame + 8 v3 views in one call |
|
| Saved object: first frame + optional 8 views |
|
| Native presets, beta dimension rules, supported controls (free) | |
| Provisional cost split into first-image + rotation units (free) |
|
Characters & Objects
Tool | Description | Key Options |
| Character via v3 model (32-256px, newest) |
|
| Character/object via Pro engine (32-168px) |
|
| Character with N/S/E/W views |
|
| Character with 8 directional views | Same as 4dir |
| New state/variant of a saved character |
|
| Animate saved character (template only) |
|
| Animate saved character (template/v3/pro) |
|
| Object, single direction (32-256px) |
|
| Object with 8 directional views (32-256px) |
|
| Animate a saved object |
|
| New state/variant of a saved object |
|
| Keep specific candidate frames as objects |
|
| Accept an object as-is (clear review) |
|
| List with pagination |
|
| Get details by ID | |
| Delete by ID | |
| Delete animations (all, or scoped) |
|
| Export character as ZIP (saved to | |
| Export as one uniform-grid spritesheet PNG + layout JSON (ZIP saved to | |
| Manage tags | |
| Attach a bust portrait to a character (free; used by |
|
Animation
Tool | Description | Key Options |
| Animate from text + reference |
|
| Animate existing image (32-256px) |
|
| Animate from first/last keyframes |
|
| Beta (tier 1+): fluid 4–40 frame clips via PixMiniMax, up to 256px |
|
| Pose control via keypoints |
|
| Edit animation frames (2-16) |
|
| Generate in-between frames |
|
| Apply outfit to frames |
|
| Extract keypoints from image |
|
Rotation
Tool | Description | Key Options |
| 8 directional views (32-168px) |
|
| 8 rotations from one frame (v3) |
|
| Rotate between views/directions |
|
Editing & Inpainting
Tool | Description | Key Options |
| Batch edit 1-16 images |
|
| Edit single image |
|
| Edit on the Pixen model; preserves pose and pixel style (max 256px source) |
|
| Mask-based editing |
|
| Inpainting (legacy, max 200px) |
|
Image Operations
Tool | Description | Key Options |
| Convert photo to pixel art |
|
| Convert photo to pixel art (Pro, auto-sizes) |
|
| AI-powered pixel art resize |
|
| Remove background (max 400px) |
|
| Recover native-resolution art from an upscaled image (min 256px, result is opaque) |
|
| Clean stray/anti-aliased pixels without resizing (batch frames together) |
|
| Quantize frames onto one shared palette, optional dithering |
|
Tilesets
Tool | Description | Key Options |
| Top-down tileset (16 or 32px) |
|
| Platformer tileset |
|
| Isometric tile (16-64px) |
|
| Pro tiles (hex, iso, octagon, square) |
|
| Retrieve by ID | |
| List with pagination |
|
| Delete by ID |
Map Objects
Tool | Description | Key Options |
| Game-ready object |
|
| Status + metadata by ID |
|
UI Assets & Fonts
Tool | Description | Key Options |
| Persistent shape-based UI panel (distinct from the one-shot |
|
| Retrieve, list (paginated), or delete UI assets |
|
| Styled pixel-art font (glyph atlas + |
|
| Convert between a bust portrait and a full-body sprite (both directions) |
|
Talking Animation
Generate mouth positions once per expression (vocal_animation is the only step that costs generations), then produce unlimited talking GIFs or engine-ready lip-sync plans for free.
Tool | Description | Key Options |
| Generate mouth positions ("visemes") for a portrait (costs generations, once per expression) |
|
| Poll a |
|
| Text → animated GIF of the character speaking (free) |
|
| Frame-by-frame mouth plan for driving lips in a game engine (free, nothing rendered) |
|
Prompt Enhancement
Expand a short description into a richer prompt. These return enhanced text only — they do not generate images.
Tool | Description | Key Options |
| Enrich a prompt for |
|
| Enrich a motion prompt for |
|
| Enrich a prompt for |
|
Account & Jobs
Tool | Description |
| Check your credit balance |
| Check a background job by ID |
| Check a font-pro / portrait↔character job (dedicated endpoints) |
| List jobs that haven't completed (for recovery after disconnection) |
| Recent job history (auto-pruned after 24h) |
| Load a saved PNG as a Base64 image to pass into other tools |
Common Options
Most tools share these parameters, but the field names differ between endpoint generations:
Concept | v2 endpoints | Legacy endpoints |
Prompt adherence | n/a |
|
Transparent background |
|
|
Color reference | n/a |
|
Style controls | n/a |
|
Negative prompt | n/a |
|
Isometric mode | n/a |
|
Reproducibility |
|
|
Character, object, tileset, and map object endpoints also accept text_guidance_scale, outline, shading, detail, and color_image.
Usage
Just describe what you want in plain language. The assistant picks the right tool and parameters automatically.
Quick Examples
Sprites and icons:
"Generate a 64x64 pixel art knight with a blue cape, no background"
"Make a 16x16 health potion icon"
"Create a 128x128 dragon boss with detailed shading and thick outlines"Characters (persistent, multi-directional):
"Create a 48x48 character with 4 directions: a dwarf blacksmith in a leather apron, chibi proportions"
"Animate that character with the walk template"
"Now add a fireball animation"Tilesets:
"Create a 32x32 top-down tileset: ocean water below, sandy beach on top, foam transition"
"Make a 16x16 sidescroller tileset with stone platforms"Editing existing art:
"Edit this sprite to make the armor gold instead of silver"
"Remove the background from this image"
"Generate 8 rotations of this character"Which Tool Gets Used?
The assistant picks the right tool automatically, but the key decision is:
generate_image— default for most requests, highest quality, largest sizesgenerate_with_style— when you want new art matching existing art ("in the same style as these sprites")generate_ui— for game UI elements (buttons, panels, health bars, icons)create_tileset/create_tiles_pro— for tileable terrain; standard for square RPG tiles, pro for hex/isometric/octagoncreate_character_4dir/8dir— for persistent characters you can animate later by ID
Key Concepts
Sizes are in pixels as
width x height— different tools have different limits (e.g. characters max 128x128,generate_imagegoes up to 792x688)Transparent backgrounds are the default on most tools — ask for a background explicitly if you want one
Characters are persistent — once created, you can animate them by ID without re-describing
Seeds make results reproducible — same seed + same description = same output
For detailed size limits, style controls, endpoint comparison, and step-by-step workflows, see the Prompting Guide.
Prompt Commands
PixelLab Forge includes MCP prompt templates that appear as slash commands in supported clients (Claude Desktop, Cursor, etc.):
Command | Description |
| Overview of all tools and how to use them |
| Generate a pixel art sprite |
| Create a character with directional views + animation |
| Animate an existing sprite or character |
| Create a top-down or sidescroller tileset |
| Create hex, isometric, or octagon tiles |
| Generate game UI elements |
| Generate art matching existing sprites' style |
| Edit or modify an existing sprite |
Reliability
Auto-polling: Generation jobs are polled every 2 seconds for up to 10 minutes
Retry on failure: Network errors during polling are retried 3 times with backoff
Job recovery: If the connection drops, job IDs are logged to stderr and persisted to disk. Use
list_pending_jobsto find them andget_job_statusto retrieve resultsImage saving: Generated images are automatically saved as PNGs to
./pixellab-forge-output/
Testing
npx @modelcontextprotocol/inspector node dist/index.jsSet PIXELLAB_API_KEY in the inspector's environment variables, connect, and test any tool.
Contributing
Contributions are welcome via pull requests.
See CHANGELOG.md for release history. Maintainers: see RELEASING.md for the release process.
License
MIT
Available Tools
103 toolsanimate_characterB
Animate a SAVED character by ID using a preset animation template (walk, run, attack, etc). Requires a character_id from create_character_4dir/8dir. Uses the character's existing size. If you only have an image (not a saved character), use animate_with_text_v2 instead.
Inputs: character_id (required), template_animation_id (required), seed, detail, +10 more.
animate_objectA
Animate a SAVED object by ID. mode='v3' (default, higher quality, supports custom_start_frame/end_frame interpolation) or 'pro'. Do NOT pass directions for 1-direction objects; for 8-direction objects pass the directions to animate. Requires an object_id from create_object_1dir/8dir.
Inputs: object_id (required), mode, end_frame, directions, +7 more.
animate_pixminimaxA
Beta (tier 1+ subscription): animate a single frame from a text description of the MOTION using the PixMiniMax engine (MiniMax H3). Best for fluid, longer clips — 4 to 40 frames (multiples of 4) at up to 256x256. Result holds frame_count + 1 images: index 0 is your unchanged input frame. Optional last_frame makes it a keyframe interpolation. Priced by clip length and size (about 1-12 generations); sizes just above 64 or 128px are cheaper than just below. Typically takes 1-5 minutes. Use animate_with_text_v3 for the standard engine.
Inputs: description (required), first_frame (required), seed, direction, +5 more.
animate_with_skeletonA
Create animation using skeleton keypoints for precise joint/pose control. Fixed sizes only: 16, 32, 64, 128, or 256px. Use this when you need exact body positioning per frame. For simpler text-described animation use animate_with_text_v2.
Inputs: view (required), direction (required), image_size (required), skeleton_keypoints (required), +10 more.
animate_with_textA
Legacy animation from text description + reference image. FIXED 64x64 only. Prefer animate_with_text_v2 (any size 32-256px) or animate_with_text_v3 (keyframe-based) for new work. Use animate_character if you already have a saved character ID.
Inputs: action (required), image_size (required), description (required), reference_image (required), +13 more.
animate_with_text_v2A
RECOMMENDED animation tool — animate a character image with a text-described action. Provide a reference image + action text. 32-256px. Frames by size: 32-64px → 16, 128-256px → 4. Use this when you have a character image but NOT a saved character ID. If you have a character ID, use animate_character instead.
Inputs: action (required), image_size (required), reference_image (required), reference_image_size (required), +4 more.
animate_with_text_v3A
Keyframe-based animation — provide a first frame and optional last frame, get interpolated animation. Best for precise start/end control. 4-16 frames, max 256px. Pixel budget: W×H×frames ≤ 524,288. Use animate_with_text_v2 instead if you just want to describe an action.
Inputs: action (required), first_frame (required), seed, last_frame, +2 more.
correct_pixelartA
Clean up existing pixel art WITHOUT changing its size: sharpens edges, removes stray/anti-aliased pixels, and tightens the palette while keeping the sprite on its pixel grid. Good for art that went through a lossy pipeline, hand-drawn art, or nearly-on-grid output from other tools. Pass several same-size frames together (an animation, or a character's directions) so they stay consistent. Max area 1024x1024; transparency is preserved. If the image is upscaled, run unzoom first.
Inputs: images (required), strength.
create_character_4dirA
Create a PERSISTENT game character with 4 directional views (N/S/E/W). 32-168px. Returns a character_id that can be reused with animate_character for animations. Use this for game characters that need multiple directions and animations. For 8 directions use create_character_8dir. For one-off sprites without persistence use generate_image.
Inputs: image_size (required), description (required), seed, view, +9 more.
create_character_8dirA
Create a PERSISTENT game character with 8 directional views (N/NE/E/SE/S/SW/W/NW). 32-168px. Returns a character_id for use with animate_character. Use this for top-down or isometric games needing diagonal directions. For 4-direction games use create_character_4dir.
Inputs: image_size (required), description (required), seed, view, +9 more.
create_character_animationA
Animate a SAVED character by ID with flexible modes. mode='template' uses a skeleton-based template_animation_id; mode='v3' interpolates frames from a custom action_description (4-16 frames); mode='pro' is the pro animator. More flexible than animate_character, which is template-only. Requires a character_id.
Inputs: character_id (required), mode, seed, detail, +13 more.
create_character_proB
Create a character or object using the Pro engine. Output 32-168px. Methods: create_with_style (text-driven, optional style reference_image), create_from_concept (from a concept_image up to 1024x1024), rotate_character. Returns a character_id.
Inputs: image_size (required), description (required), seed, view, +6 more.
create_character_pro_flashA
Create a saved character with the Pro Flash engine in one call: generates a south-facing image from the description, then produces eight v3 rotation views. To skip the (paid) first image, pass source_image_id from an earlier create_image_pro_flash result, or upload first_frame directly — you then pay only for rotations. Pixels are preserved and padded with transparency to the square v3 canvas. Returns a job plus character_id; poll get_job_status for itemized billing_stages, then get_character.
Inputs: description (required), name, seed, view, +8 more.
create_character_stateA
Create a new state/variant of an existing character by editing it with a text description (e.g. 'wearing armor', 'damaged'). Requires a source character_id. Optionally snap the edit to the source character's color palette for visual consistency.
Inputs: character_id (required), edit_description (required), seed, no_background, +1 more.
create_character_v3A
Create a character using the v3 model. Two modes: provide a south-facing reference_image to rotate that exact character, or omit it to generate from scratch via description. Output 32-256px (advisory in reference mode). Returns a character_id. The newest character generator — prefer over create_character_4dir/8dir for best quality.
Inputs: description (required), name, seed, view, +7 more.
create_image_bitforgeA
Generate pixel art using the legacy Bitforge engine. Max 200x200. Prefer generate_image (v2) for most tasks — use Bitforge only when you need inline skeleton keypoints, combined inpainting+generation, or style_image influence control in a single call.
Inputs: image_size (required), description (required), seed, view, +20 more.
create_image_pixenA
Generate pixel art using the Pixen engine. Width/height 16-768px each but max area 512x512, and both must be divisible by 4. Supports outline/detail style hints, view/direction, and optional prompt enhancement. A modern alternative to generate_image with finer outline/detail control.
Inputs: image_size (required), description (required), seed, view, +6 more.
create_image_pixfluxA
Generate pixel art using the legacy Pixflux engine. 32-400px. Prefer generate_image (v2) for most tasks — use Pixflux only when you need fine-grained control over outline/shading/detail style params, view/direction, or coverage_percentage.
Inputs: image_size (required), description (required), seed, view, +12 more.
create_image_pixflux_backgroundA
Generate a seamless/background image using the Pixflux engine. Same parameters as create_image_pixflux but tuned for backgrounds and environments rather than isolated sprites.
Inputs: image_size (required), description (required), seed, view, +12 more.
create_image_pro_flashA
Generate one native pixel-art image with the Pro Flash engine, optionally guided by a style image. Runs as a background job; the completed result carries the PNG, an image_url and a durable source_image_id — keep that ID to later build a Pro Flash character or object from the same image without paying for it again (create_character_pro_flash / create_object_pro_flash). About 5 generation units; call get_pro_flash_cost for current pricing and get_pro_flash_capabilities for the exact native sizes.
Inputs: image_size (required), description (required), seed, project_id, +3 more.
create_isometric_tileA
Create an isometric tile. Size 16x16 to 64x64 (best quality >24px). For other tile types (hex, octagon, square_topdown) use create_tiles_pro instead.
Inputs: image_size (required), description (required), seed, detail, +8 more.
create_map_objectA
Generate a single-view map object (tree, rock, chest, etc.) with transparent background. Use this for environment/prop assets that don't need multiple directions. For objects with 4 directional views use create_object_4dir. For characters use create_character_4dir/8dir.
Inputs: image_size (required), description (required), seed, view, +9 more.
create_object_1dirA
Create a persistent single-direction object. Square size 32-256px (default 64). view 'top-down' or 'sidescroller'. Optionally pass style_images for visual reference. Larger sizes may yield multiple objects (use item_descriptions to label each). For an object rotatable to 8 directions use create_object_8dir. For single-view non-persistent props use create_map_object.
Inputs: description (required), size, view, style_images, +1 more.
create_object_8dirA
Create a persistent object with 8 directional views. Square size 32-256px (default 64). view 'low top-down', 'high top-down', or 'side'. Provide reference_image to rotate that exact object, OR style_image to generate a new object in that style (mutually exclusive). For a single-direction object use create_object_1dir.
Inputs: description (required), size, view, style_image, +1 more.
create_object_pro_flashA
Create a saved object with the Pro Flash engine in one call: generates a south-facing image from the description and, with n_directions 8, adds v3 rotation views (n_directions 1 = single image). Pass source_image_id or first_frame to reuse an existing image: one-direction finalization is then free and eight views charge only rotations. Objects have no skeleton. Returns a job plus object_id; poll get_job_status for billing_stages, then get_object.
Inputs: description (required), name, seed, view, +7 more.
create_object_stateB
Create a new state/variant of an existing object by editing it with a text description (e.g. 'open chest', 'broken'). Requires a source object_id.
Inputs: object_id (required), edit_description (required), seed.
create_tilesetA
Create a TOP-DOWN tileset with base terrain, elevated terrain, and transitions (16x16 or 32x32 tiles). Outputs 16-23 seamless tiles. Use this for RPG/strategy maps. For platformer/sidescroller games use create_tileset_sidescroller. For isometric games use create_isometric_tile or create_tiles_pro.
Inputs: tile_size (required), lower_description (required), upper_description (required), seed, +16 more.
create_tileset_sidescrollerA
Create a SIDESCROLLER/PLATFORMER tileset with terrain and transitions (16x16 or 32x32 tiles). Side-view perspective only. Use this for platformer games. For top-down RPG maps use create_tileset instead.
Inputs: tile_size (required), lower_description (required), seed, detail, +12 more.
create_tiles_proA
Create professional tiles. Types: hex, hex_pointy, isometric, octagon, square_topdown. Size 16-128px (32px recommended).
Inputs: n_tiles (required), tile_size (required), tile_type (required), description (required), +7 more.
create_ui_assetA
Generate a shape-based pixel-art UI panel (Pro) from a text description — a persistent, saved UI asset (distinct from generate_ui, which is a one-shot generator). Returns a job_id and a ui_asset_id immediately; poll get_job_status or get_ui_asset until ready. Optionally scaffold the panel from named UI elements or a custom shape template.
Inputs: description (required), name, seed, pieces, +6 more.
delete_characterC
Delete a character by ID.
Inputs: character_id (required).
delete_character_animationsA
Delete animations from a character. Omit all optional filters to delete every animation; pass animation_type and/or animation_group_id (both shown by get_character) to narrow it, and direction to remove a single direction only.
Inputs: character_id (required), direction, animation_type, animation_group_id.
delete_isometric_tileB
Delete an isometric tile by ID.
Inputs: tile_id (required).
delete_objectC
Delete an object by ID.
Inputs: object_id (required).
delete_object_animationsA
Delete animations from an object. Omit all optional filters to delete every animation; pass animation_type and/or animation_group_id (both shown by get_object) to narrow it, and direction to remove a single direction only.
Inputs: object_id (required), direction, animation_type, animation_group_id.
delete_tilesetA
Delete a top-down tileset by ID. For sidescroller tilesets use delete_tileset_sidescroller.
Inputs: tileset_id (required).
delete_tileset_sidescrollerA
Delete a sidescroller tileset by ID. For top-down tilesets use delete_tileset.
Inputs: tileset_id (required).
delete_tiles_proB
Delete a pro tile by ID.
Inputs: tile_id (required).
delete_ui_assetA
Permanently delete a UI panel and its backing image files.
Inputs: ui_asset_id (required).
dismiss_object_reviewB
Dismiss the review state on an object, accepting it as-is without selecting specific frames. Takes no body beyond the object_id.
Inputs: object_id (required).
download_character_spritesheetA
Export a character as a single spritesheet: a small ZIP holding one uniform-grid PNG (row 0 = rotations, then one row per animation-direction, columns = frames) plus a layout JSON describing cell size and which animation/direction each row holds. Frames are centred in equal cells and never rescaled. Saves to pixellab-forge-output and returns the file path. Use download_character_zip instead for individual frame PNGs. Returns an error while the character is still generating (HTTP 423).
Inputs: character_id (required).
download_character_zipA
Download a character as a ZIP file with all sprites and metadata. Saves it to the pixellab-forge-output directory and returns the file path.
Inputs: character_id (required).
download_object_spritesheetA
Export an object as a single spritesheet: a small ZIP holding one uniform-grid PNG (row 0 = rotations, then one row per animation-direction, columns = frames) plus a layout JSON describing cell size and which animation/direction each row holds. Frames are centred in equal cells and never rescaled. Saves to pixellab-forge-output and returns the file path. Only works for objects you created; errors while rotations are still generating.
Inputs: object_id (required).
edit_animationA
Edit an existing animation sequence (2-16 frames) using a text description. 16-256px. Use this to modify animations you already have — to create new animations from scratch use animate_with_text_v2 or animate_character.
Inputs: frames (required), image_size (required), description (required), seed, +1 more.
edit_imageA
Edit a single image globally using a text description (e.g. 'add a hat', 'change colors'). 16-400px. For editing a specific REGION use inpaint_v3 with a mask instead. For batch-editing multiple images at once use edit_images.
Inputs: image (required), width (required), height (required), image_size (required), +5 more.
edit_image_pixenA
Edit an existing pixel art image with a text instruction on the Pixen model — pose, composition and pixel style are preserved and only what you ask for changes. Source max 256px per side (crop, don't rescale). Target canvas defaults to the source size; area 16x16 to 256x256 (a wide-but-short 128x512 is fine). The model re-renders at the target size rather than rescaling. Costs 1 generation. Prefer this over edit_image for Pixen-generated art.
Inputs: image (required), description (required), seed, width, +2 more.
edit_image_pro_flashA
Edit one image at a supported native size with the Pro Flash engine, using either a text instruction (method 'text') or a reference image (method 'reference'). The source canvas never grows or resizes, and reference art must fit inside it. Runs as a background job. Call get_pro_flash_capabilities for the exact native shapes.
Inputs: image (required), seed, method, project_id, +4 more.
edit_imagesA
Batch-edit 1-16 images consistently using text or a reference image. Use this for editing animation frames or sprite sets uniformly. Output 32-512px, input max 256px each. Max frames by output size: 32-64px → 16, 65-80px → 9, 81-128px → 4, 129-512px → 1. For single image edits use edit_image instead.
Inputs: method (required), image_size (required), edit_images (required), seed, +3 more.
enhance_animation_promptA
Expand a short action description into a richer motion prompt for animate_with_text_v3. Provide the first frame (and optional last frame for interpolation). Returns enhanced text — does not generate frames.
Inputs: action (required), first_frame (required), last_frame.
enhance_character_promptA
Expand a short character description into a richer, more detailed prompt for create_character_v3. Returns enhanced text — does not generate an image. Use this to preview/refine a prompt before generating.
Inputs: image_size (required), description (required), view, detail, +1 more.
enhance_pixen_promptA
Expand a short image description into a richer prompt for create_image_pixen. Returns enhanced text — does not generate an image.
Inputs: image_size (required), description (required), view, detail, +3 more.
estimate_skeletonB
Estimate skeleton keypoints from a character image.
Inputs: image (required).
generate_8_rotationsA
Generate 8 directional views from an existing image, style, or concept art. 32-168px. Does NOT create a persistent character — for that use create_character_8dir instead. Methods: rotate_character (from existing sprite), create_with_style (from description), create_from_concept (from concept art).
Inputs: method (required), image_size (required), seed, view, +5 more.
generate_8_rotations_v3A
Generate 8 directional rotations from a single source frame using the v3 model. Simpler than generate_8_rotations — just provide one image and the model produces all 8 views. Does NOT create a persistent character; use create_character_8dir for that.
Inputs: first_frame (required), seed, no_background.
generate_font_proA
Generate a styled pixel-art font (Pro) from a text description. Produces a glyph atlas plus a ready-to-use TrueType (.ttf) font. Returns a job_id immediately — poll get_job_status for the result.
Inputs: weight (required), description (required), seed, glyph_px, +2 more.
generate_imageA
Generate pixel art from a text description. PRIMARY generation tool — use this for standalone sprites, icons, and one-off images. 16-512px. Variants by size: ≤42px → 64, 43-85px → 16, 86-170px → 4, >170px → 1. For style-matched sets use generate_with_style. For game characters with directional views use create_character_4dir/8dir instead. For UI elements use generate_ui.
Inputs: image_size (required), description (required), seed, style_image, +3 more.
generate_uiA
Generate pixel art UI elements for games — buttons, panels, health bars, inventory slots, icons, frames. Use this instead of generate_image when creating interface/HUD elements. Min 16x16, max 512x512.
Inputs: image_size (required), description (required), seed, color_palette, +2 more.
generate_with_styleA
Generate pixel art matching a specific visual style from 1-4 style reference images. Use this when you need consistent style across multiple assets. SQUARE images only, 16-512px. Auto-pads to nearest bucket (16/32/64/128/256/512). Variants: 16-32px → 64, 33-64px → 16, 65-128px → 4, 129-512px → 1.
Inputs: image_size (required), description (required), style_images (required), seed, +2 more.
get_balanceA
Get your current PixelLab credit balance
Inputs: no named parameters.
get_characterA
Get a character by ID including all directional views and animations.
Inputs: character_id (required).
get_font_pro_jobA
Get the status and result of a generate_font_pro job by its job_id. Use this rather than get_job_status for font-pro jobs — they are served from a dedicated endpoint.
Inputs: job_id (required).
get_isometric_tileA
Get a previously created isometric tile by ID.
Inputs: tile_id (required).
get_job_statusA
Check the status of a background job and retrieve its results when complete. All creation tools return a job_id immediately — use this tool to poll for completion and get the generated images/data.
Inputs: job_id (required).
get_map_objectA
Get a map object's status and metadata by ID (from create_map_object).
Inputs: object_id (required).
get_objectC
Get an object by ID.
Inputs: object_id (required).
get_portrait_character_pro_jobA
Get the status and result of a portrait_character_pro job by its job_id. Use this rather than get_job_status for portrait↔character jobs — they are served from a dedicated endpoint.
Inputs: job_id (required).
get_pro_flash_capabilitiesA
List the Pro Flash engine's native size presets, beta custom-dimension rules, and supported controls. Free. Check this before calling create/edit/inpaint_image_pro_flash with a non-standard size.
Inputs: no named parameters.
get_pro_flash_costA
Estimate the provisional cost of a Pro Flash operation in generation units, split into the first-image component and the v3 rotation component. Free. Actual billed usage is reported when the job completes.
Inputs: width (required), height (required), operation (required), n_directions.
get_tilesetB
Get a previously created tileset by ID.
Inputs: tileset_id (required).
get_tileset_sidescrollerA
Get a previously created sidescroller tileset by ID. For top-down tilesets use get_tileset.
Inputs: tileset_id (required).
get_tiles_proB
Get previously created pro tiles by ID.
Inputs: tile_id (required).
get_ui_assetB
Get a UI panel's details by ID. Reports progress while the panel is still generating.
Inputs: ui_asset_id (required).
get_vocal_animation_jobA
Get the status and result of a vocal_animation job by its job_id. Mouth positions stream in as they are produced (completed_visemes fills up while the job runs). On completion, a character_id job has saved the set onto the character; a portrait job returns the frames in visemes.
Inputs: job_id (required).
image_to_pixelartA
Convert a photograph or regular image into pixel art. Input 16-1280px, output 16-320px. Recommended: output = 1/4 of input size.
Inputs: image (required), image_size (required), output_size (required), seed, +1 more.
image_to_pixelart_proA
Convert a photograph or regular image into pixel art using the Pro engine. The model derives output size from the input — you only supply the source image and optional style instructions. Higher quality than image_to_pixelart but less manual size control.
Inputs: image (required), seed, description.
inpaintA
Legacy inpainting using the Bitforge engine. Max 200x200. Prefer inpaint_v3 (up to 512px, better quality). Only use this if you need Bitforge-specific params like style controls or oblique projection during inpainting.
Inputs: image_size (required), mask_image (required), description (required), inpainting_image (required), +15 more.
inpaint_image_pro_flashA
Pro Flash inpainting: replace only the WHITE pixels of mask_image, preserving unmasked RGBA exactly. Source and mask must share identical supported native dimensions; mask RGB must be pure black/white (alpha is ignored) and may not be empty. output_method controls whether you get the full composite or only the changed pixels (transparent elsewhere). An optional context_image requires its bounding_box in the source. Runs as a background job.
Inputs: image (required), mask_image (required), description (required), seed, +7 more.
inpaint_v3A
Edit a specific region of a pixel art image using a mask. White mask = generate, black mask = preserve. Image 32-512px, optional context image up to 1024x1024. Preferred over legacy inpaint (which caps at 200px).
Inputs: mask_image (required), description (required), inpainting_image (required), seed, +4 more.
interpolate_framesA
Generate intermediate animation frames between a start and end keyframe. 16-128px. Use this to smooth out animations by adding in-between frames. For full animation creation from a single image use animate_with_text_v2.
Inputs: action (required), end_image (required), image_size (required), start_image (required), +2 more.
lip_syncA
Get the frame-by-frame lip-sync plan for a line of text — which mouth position to show, for how long, and how far through the text it lands. Free, and nothing is rendered: use this instead of talking_gif when animating in a game engine and driving the mouth yourself. Provide character_id (response also carries the spritesheet URL and row to read) or a bare viseme_count preset.
Inputs: text (required), mood, hold_ms, frame_ms, +2 more.
list_charactersA
List your created characters with pagination.
Inputs: limit, offset.
list_isometric_tilesB
List your previously created isometric tiles with pagination.
Inputs: limit, offset.
list_job_historyA
List recent job history (completed, failed, and pending). Jobs are pruned after 24 hours.
Inputs: no named parameters.
list_objectsC
List your created objects with pagination.
Inputs: limit, offset.
list_pending_jobsA
List background jobs that were started but haven't completed yet. Use this to recover jobs after a disconnection or timeout.
Inputs: no named parameters.
list_tilesetsB
List your previously created tilesets with pagination.
Inputs: limit, offset.
list_tilesets_sidescrollerA
List your sidescroller tilesets with pagination. For top-down tilesets use list_tilesets.
Inputs: limit, offset.
list_tiles_proB
List your previously created pro tiles with pagination.
Inputs: limit, offset.
list_ui_assetsA
List your saved UI panels (newest first), with pagination. Includes ghost rows for panels still generating.
Inputs: limit, offset.
portrait_character_proA
Convert between a bust portrait and a full-body character sprite (Pro). direction='portrait_to_character' takes a portrait in and returns a full-body sprite; 'character_to_portrait' does the reverse. Returns a job_id immediately — poll get_job_status for the result.
Inputs: image (required), seed, view, direction, +1 more.
read_imageA
Read a previously saved image from disk and return it as a Base64Image object that can be passed directly to other tools (e.g. edit_image, remove_background, image_to_pixelart). Use the file paths shown in earlier tool responses.
Inputs: file_path (required).
reduce_colorsA
Quantize one or more images onto a smaller SHARED palette, optionally with ordered dithering. Pass an animation's frames or a character's eight directions in one call so they come back sharing one palette instead of drifting apart. Choose the palette by omitting both options (auto-detect size), num_colors (exact count), or palette_image (reuse an existing image's colors, max 256). Total pixel budget across all frames is 512x512 (e.g. sixteen 64x64 frames).
Inputs: images (required), dithering, num_colors, palette_image, +1 more.
remove_backgroundC
Remove the background from a pixel art image (max 400x400).
Inputs: image (required), image_size (required), seed, text, +1 more.
resize_imageA
AI-powered resize of a pixel art image to a different resolution while preserving quality. Source and target 16-200px. Best in steps: max 50% shrink or 2x grow per operation.
Inputs: description (required), target_size (required), reference_image (required), reference_image_size (required), +9 more.
rotateA
Rotate a single character sprite from one view/direction to another. Fixed sizes only: 16, 32, 64, or 128px. For generating all 8 directions at once use generate_8_rotations or create_character_8dir instead.
Inputs: from_image (required), image_size (required), seed, to_view, +12 more.
select_object_framesB
From an object that generated multiple candidate frames, keep specific frames (by 0-based index) as completed individual objects. Optionally tag all newly-created objects.
Inputs: indices (required), object_id (required), common_tag.
set_character_portraitA
Attach a bust portrait to a saved character (free — no generation runs). The portrait is the starting frame for talking animations: vocal_animation generates mouth positions from it. Overwrites any existing portrait. To generate a portrait from a full-body sprite first, use portrait_character_pro with direction='character_to_portrait'.
Inputs: image (required), character_id (required).
talking_gifA
Turn a line of text into an animated GIF of a character speaking it. Free — spends no generations; it only re-orders mouth positions already produced by vocal_animation. Provide either character_id (with mouth positions stored on the character) or supply visemes directly as returned by get_vocal_animation_job.
Inputs: text (required), mood, hold_ms, visemes, +2 more.
transfer_outfitA
Transfer an outfit/costume from a reference image onto animation frames (2-16 frames, 32-256px). Use this to reskin an existing animation with a different character appearance.
Inputs: frames (required), image_size (required), reference_image (required), seed, +1 more.
unzoomA
Recover native-resolution pixel art from an upscaled image (e.g. a 32x32 sprite saved at 512x512). Detects the underlying pixel grid and downsamples back onto it, handling slightly uneven grids from lossy resizes. Input min 256x256, max area 2048x2048. The result is OPAQUE (transparency is composited onto white) — run remove_background afterwards if you need it cut out again. Use before passing user-supplied art as a style/reference image, or before correct_pixelart.
Inputs: image (required), quantize.
update_character_tagsA
Update tags on a character (max 20 tags, 50 chars each).
Inputs: tags (required), character_id (required).
update_object_tagsC
Update tags on an object.
Inputs: tags (required), object_id (required).
vocal_animationA
Generate the set of mouth positions ('visemes') that lets a portrait be lip-synced to any line of text. This is the only talking-animation step that costs generations — pay it once per expression, then talking_gif and lip_sync are free and unlimited. Provide either character_id (uses the character's stored portrait, set via set_character_portrait, and saves the result onto it) or an inline portrait image (max 256x256, result returned inline). Returns a job_id — poll get_vocal_animation_job.
Inputs: mood, seed, portrait, character_id, +2 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 103 tools
Many tools have near-overlapping purposes: generate_image, create_image_bitforge, create_image_pixflux, create_image_pixen, and create_image_pro_flash all generate pixel art, and character/animation creation is similarly fragmented across 4dir/8dir/v3/pro/pro_flash and v2/v3 text animation variants. The descriptions try to steer selection, but the boundaries are still unclear enough that an agent will frequently face ambiguous choices.
Most names follow a readable verb_noun snake_case pattern, but consistency breaks down with irregular version/engine suffixes like v2, v3, pro, pro_flash, bitforge, pixflux, and pixen pasted onto similar operations. The broader pattern is recognizable, but tool-to-tool naming is not predictable enough to guess the right variant.
103 tools is an extreme count for a single MCP server, driven largely by multiple overlapping engine generations and legacy alternatives for the same core capabilities. This is well beyond a manageable scope and will impose a heavy navigation burden on agents.
Domain coverage is very strong: characters, objects, tilesets, UI assets, animations, images, jobs, and talking portraits all have solid create/read/update/delete and conversion workflows. Minor gaps exist—fonts have generate/get but no list/delete, and map objects lack full lifecycle management—but no critical dead ends block the main pixel-art generation workflow.
Maintenance
Related MCP Connectors
Pixel art for your game, from your coding agent. Generate sprites, items and tiles in your game's style (64 to 1024 px), make variants, reskins and N/E/S/W directions, then edit everything in a full pixel art editor that syncs into your project. Works with Claude, Claude Code, Cursor and ChatGPT; sign in with OAuth, no API key.
Generate authentic pixel art - sprites, animations, and tilesets - from any MCP client
MCP server for AI dialogue using various LLM models via AceDataCloud
MCP server for Flux AI image generation
Related MCP Servers
- AlicenseAqualityBmaintenanceMCP server that forges pixel art sprites & game assets using Google Gemini — generate, crop, split & process, all from your AI assistant.68149 npm6MIT
- AlicenseNot gradedqualityDmaintenanceMCP server for programmatically creating and editing Aseprite sprites, enabling AI agents to draw, manage layers and frames, and iterate until the desired result is achieved.MIT
- AlicenseBqualityDmaintenanceMCP server for Aseprite — create, edit, and export pixel art sprites, animations, and sprite sheets from any AI assistant.438MIT
- AlicenseNot gradedqualityFmaintenanceMCP server for generating style-consistent pixel art assets from PixelLab API, with project management, asset review, and an embedded Claude assistant.MIT