@textile-designer-ai/mcp
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., "@@textile-designer-ai/mcp make a half-drop repeat of paisley.png at 300 dpi and tell me the cost"
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.
@textile-designer-ai/mcp
MCP server for Textile Designer AI. It lets Claude Code, Claude Desktop, Codex, Cursor or any other MCP client run the Textile Designer tools on image files: upscale, repeat, background removal, channeling, colourways, vectorizer and the rest.
It is a thin client of https://www.textile-designer.ai/api/plugin/*, the
same API the Photoshop plugins use. You log in with your Textile Designer
account; credits, plan access and enabled tools are decided by the server.
Photoshop is not involved: inputs are image files or URLs, outputs are files.
Requires Node.js 18 or newer.
About this repository
This repository holds the published build of the MCP server (dist/index.js, a single file with no runtime dependencies), exactly as released on npm as @textile-designer-ai/mcp. Run it with npx -y @textile-designer-ai/mcp, or node dist/index.js from a clone. A Dockerfile is included for registries that inspect the server in a container. The Claude plugin bundle lives in ScientiaAI/textile-designer-ai-claude.
Security reports: https://www.textile-designer.ai/contact (subject "Security").
Related MCP server: Ads Optimiser MCP server
Install
Claude Code
claude mcp add textile-designer-ai -- npx -y @textile-designer-ai/mcpClaude Desktop
Add to claude_desktop_config.json (Settings → Developer → Edit Config):
{
"mcpServers": {
"textile-designer-ai": {
"command": "npx",
"args": ["-y", "@textile-designer-ai/mcp"]
}
}
}Codex
Add to ~/.codex/config.toml:
[mcp_servers.textile_designer_ai]
command = "npx"
args = ["-y", "@textile-designer-ai/mcp"]Cursor
Add to .cursor/mcp.json (project) or ~/.cursor/mcp.json (global):
{
"mcpServers": {
"textile-designer-ai": {
"command": "npx",
"args": ["-y", "@textile-designer-ai/mcp"]
}
}
}On Windows, if npx is not found by the client, use "command": "npx.cmd"
or the full path to it.
Logging in
Ask the assistant to connect, or call the login tool. The server starts a
device-code login: it opens textile-designer.ai/plugin-auth in your browser
with a short code, you confirm it there, and the server picks up the session
in the background. whoami confirms the connection and shows your credits,
plan access and enabled tools. logout ends the session.
The session token is stored in ~/.textile-designer/mcp-session.json
(readable only by you). It is never printed.
Example conversation:
Connect to Textile Designer, then make a half-drop repeat of
C:\designs\paisley.pngat 300 dpi and tell me what it cost.
Tools
Account and jobs:
Tool | What it does |
| Start the device-code login (opens the browser; |
| Email, organisation, credits, plan access, enabled tools. |
| End the session and forget the token. |
| Status / progress of a job by |
| Save a completed job's outputs to disk. |
| Every design tool with a one-line description and its credit rule. |
| One tool's best use, how it compares with similar tools, what its settings change, its outputs, modes and credit rule. |
| Answers "which tool should I use", typical workflows (digital, screen or rotary, garment to collection, vector), comparisons, print basics, results, credits and plans. |
| Truly Unlimited only: use a Boost to double today's budget and start queued slow-lane jobs now. Claude asks you first; it spends a boost. |
Design tools (website names):
Tool | Website label | Result |
| Ready to Print | image |
| Anti-Blur | image |
| Super Scaler | image |
| Background Removal | image |
| Watermark Removal | image |
| Repeat Set | image |
| Design Extension | image |
| Border Outline | image |
| Object Layering | PSD |
| Colour Layering | PSD |
| Style Transfer (library style or | image |
| Dress to Design | 1-4 images |
| Channeling | multichannel PSD + preview |
| Color Transfer (needs | image |
| Color Matching: extract palette | JSON |
| Color Matching: Suggest Colors (market, palette type) | JSON + ready-made replace sets |
| Color Matching: match colours from reference images | JSON + replace sets |
| Color Matching Auto, step 1 (extract colours first) | grid image |
| Color Matching Auto, step 2 | images |
| Color Matching: Replace Color | image(s) |
| Vectorizer (Photoshop-Ready or Illustrator-Ready) | png / tiff / eps (Photoshop-Ready), svg / pdf / eps (Illustrator-Ready) |
| 3D Effect | images |
| Embroidery Effect | image |
| Fabric Texture Removal | images |
| Sketch to Design | image |
| Foil Separation | image / eps |
| Design Generation (Image+Text, Image to Image, Inpaint) | image |
| Design Creation (prime / new / old / reference) | images |
| Colourways | buyer sheet or images |
| Spec Board | image |
Color Matching (FREE) / psd_variant is not available: it runs inside
Photoshop and has no server job.
Every design tool accepts:
image: local path or http(s) URL. PNG, JPEG or WebP, up to 80 MB and 5000 px per side.output_dir: where to save. Default: the input image's folder (the current directory for URL inputs). Existing files are never overwritten.dpi(tools with a DPI control on the website): output resolution 72-500. Omit to keep the image's own resolution.review_settings: when you give no mode or settings, the tool asks first: the mode, then "use the recommended settings, or adjust them one by one?".truegoes straight to each setting;falseruns with the given or recommended settings (Claude passes it after you chose).unlimited_bundle: request the Unlimited bundle for this run (Unlimited plans only; the server decides).estimate_only: return the credit estimate without uploading or charging.wait(default true) andtimeout_seconds(default 240): wait for the job and save the outputs. If the job takes longer, the tool returns thetask_idand statusprocessing; usejob_statusthendownload_result.wait: falsereturns right after submit.
Tool-specific parameters are exactly the controls the website form shows
(same labels, options, ranges and recommended values), described in each
tool's schema. Creativity is shown as a percentage, as on the website, and
sent as a decimal. Nothing runs until the mode and settings are settled, and
prices are quoted only when you ask (estimate_only).
Truly Unlimited plans
One image per run: the number-of-outputs question is skipped and 1 is sent (Dress to Design All Modes still returns one image per mode).
Super Scaler 8x is not offered.
Runs over today's budget are queued in the slow lane: the tool returns at once with the expected time; check later with
job_status.whoamishows today's usage and boosts left, andboost_unlimitedstarts queued jobs now (only after you say yes).
Modes and costs
Tools with modes (Super Scaler's AI Mode, Dress to Design's extraction
models, Design Creation's variants, and so on) describe every option in their
schema as value (website label): what it does, and they refuse to run until
that mode argument is given. When the client supports MCP elicitation (a
form prompt), the server asks you directly with the website default marked
"(Recommended)" and preselected; otherwise the call returns
status: mode_required with the menu, nothing is uploaded or charged, and the
assistant is instructed to present the modes as options and let you pick.
Website defaults are documented but never applied silently.
After the mode, the server asks once whether to run with the recommended
settings or adjust them; "adjust" (or review_settings: true) then asks each
website control in turn, one question at a time, in the website's order, with
the recommended value (or the value already given) preselected. Declining any
question runs nothing. Without elicitation the result lists the same questions,
numbered, with the recommended answer marked. Two read-only tools back this up:
list_tools: every design tool with a one-line description and its credit rule, for "what can you do?".describe_tool: one tool's modes with what each option does, its settings (website controls with ranges and recommended values), its credit rule, required inputs and result kind, for "what does Prism Shift do?".
Credit figures in descriptions are the website's price list. The server's
estimate is authoritative: ask "what would this cost?" and the assistant calls
the tool with estimate_only: true, which uploads nothing and charges nothing.
After a run, the result carries the credits actually charged.
Prompts (guided workflows)
The server also publishes three MCP prompts. In Claude Code they appear as
slash commands (/textile-designer-ai:print_ready_pipeline and so on); other
clients list them under prompts. Each one fills in a message that walks the
assistant through a fixed sequence, with the rule that every paid step is
estimated and shown to you first, any unsettled mode is asked for, credits
are reported after each step, and steps are chained by the saved file paths.
Prompt | Arguments | Steps |
|
| Ready to Print -> Repeat Set -> Channeling |
|
| Dress to Design -> Colourways -> Spec Board (garment type asked) |
|
| Vectorizer (format asked from the chosen set) -> Border Outline |
Every tool also carries MCP annotations: the account, job-status and
catalogue tools are read-only; download_result writes files but charges
nothing; every td_* tool is marked as a paid, non-idempotent run.
Interactive previews (MCP Apps)
In hosts that support MCP Apps (claude.ai and Claude Desktop, ChatGPT, VS
Code, Goose, Postman as of the ext-apps 1.7 docs), every td_* result renders
as a small panel in the conversation: the saved images, their paths and the
credits charged, or the mode menu when a run still needs a mode. The Color
Matching swatch grid renders as a picker: click the cells you like and the
panel sends the chosen variant numbers back so the assistant can estimate and
run td_color_matching_regenerate. The pages are self-contained (no CDN,
no network), follow the host's light / dark theme, and are built into the
package. Hosts without MCP
Apps (Claude Code's terminal, Codex, Cursor) ignore them and get the normal
text result with the first image inline.
Suggestions
After a run completes and its files are saved, the server may add one tip line
to the result (and structuredContent.suggestions):
Image Search, when the input image's folder holds 50 or more designs.
Cataloguing, after the third Dress to Design run in a session.
Each tip appears at most once per server session, never after a failed or
estimate-only call, and never for an account that already has the product
(/me.products.<product>.active). If the server could not check Image Search
ownership (checked: false) the tip is held back. Servers that do not report
products at all get each tip once.
Results
Files are saved as <input name>_<tool>[_<n>].<ext>:
Images: one file per output. Multi-output tools that deliver a zip are unpacked into
<input name>_<tool>_<task id>/.Object / Colour Layering: a
.psd.Channeling: the separation
.psd(one spot channel per screen) unpacked from the bundle, plus<input name>_channeling_preview.png.Color Matching extract: a
.jsonpalette; the parsed colours are also in the structured result.
The tool result lists the paths, the credits charged and the task_id, and
includes the first image inline (when it is small enough) so the assistant
can look at it. Results are also kept in your Library on the website.
Environment variables
Variable | Purpose |
| API origin. Default |
| Use this session token instead of the stored one (scripted agents, CI). |
| Where to store the token. Default |
Set them in the client's MCP config (env block) when needed.
Available Tools
39 toolsboost_unlimitedBoost Truly UnlimitedAIdempotent
Truly Unlimited plans: Boost doubles today's budget and starts the runs parked in the slow lane now. It uses one of the designer's free monthly boosts or a purchased team boost. Only call it after the user explicitly says yes to using a boost in this conversation. When a boost is already on today it just starts the parked jobs (no boost spent); run_now: true asks for only that.
Inputs: run_now.
describe_toolDescribe a design toolARead-onlyIdempotent
A tool's best use, how it compares with similar tools, what its settings do to the result, its outputs, its modes with what each option does, its settings (the website controls with ranges and recommended values), its credit rule, required inputs and result kind. Call it before running a tool with modes the user has not chosen, or when the user asks what a mode does.
Inputs: tool (required).
download_resultDownload job outputsAIdempotent
Save a completed job's outputs to disk (images, unpacked zips, PSD, Channeling bundle). Use after job_status reports completed, or to fetch a job's files again.
Inputs: task_id (required), kind, name, output_dir.
guideTextile Designer AI guideARead-onlyIdempotent
Answers how-to questions from the website's own documentation: which tool fits a goal (choose_tool), typical chains (workflows), how similar tools differ (compare, optionally for one tool), DPI / repeat / format basics (print_basics), where results go and why they can differ from the original (results), estimates, Truly Unlimited and Boost (credits_and_plans), and what to say about the AI models (models).
Inputs: topic (required), tool.
job_statusJob statusARead-onlyIdempotent
Status of a submitted job (processing / completed / failed / cancelled) with progress and result URLs. Optionally wait up to wait_seconds for it to finish.
Inputs: task_id (required), wait_seconds.
list_toolsList the design toolsARead-onlyIdempotent
Every design tool with a one-line description and its credit rule. Use this to answer 'what can you do' or to pick a tool; call describe_tool for a tool's modes.
Inputs: no named parameters.
loginConnect to Textile Designer AIARead-onlyIdempotent
Start the device-code login. Returns a URL and a code; the browser opens automatically unless open_browser is false. The server keeps polling in the background: call whoami to confirm, or set wait_seconds to block until linked.
Inputs: open_browser, wait_seconds.
logoutDisconnectBRead-onlyIdempotent
End the session on the server and forget the stored token.
Inputs: no named parameters.
td_3d_effect3D EffectA
3D Effect: add depth/relief to the design. similar (1-4 outputs) or same (keeps the design fixed, one output). Best for: turning a flat 2D textile design into a 3D look with depth, shadows and embossing, e.g. showing a buyer embossed printing without a physical sample. Clear, well-defined patterns work best. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, wait, output_dir, +7 more.
td_anti_blurAnti-BlurA
Anti-Blur: sharpen a blurred design. Pro mode (recommended) or Legacy mode; files over 20 MB run in Legacy automatically. Best for: removing blur, noise and softness from a design or photo: out-of-focus or motion-blurred images, soft smartphone fabric photos, blurry social-media references. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, mode, wait, +5 more.
td_background_removalBackground RemovalA
Background Removal: remove the background and replace it with solid white. Precision or Ultra model (Unlimited on Unlimited plans). Best for: isolating a subject or motif from its background: extracting photographed motifs to build a new pattern, clean product shots, or prepping a design for layering. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, wait, model, +6 more.
td_border_outlineBorder OutlineA
Border Outline: trace the outline of the design. No settings. Best for: a clean outline-only version of a filled design: outline prints for screen printing or outline drawings for tech packs, without tracing by hand. High-contrast, well-defined art works best. Result files are saved to disk and their paths returned.
Inputs: image (required), wait, output_dir, estimate_only, +3 more.
td_channelingChannelingA
Channeling: separate the design into print screens (spot channels). Returns a multichannel PSD (one channel per screen) plus a preview PNG. Free for a limited time; priced by the server. Best for: preparing a design for screen or rotary printing. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, inks, wait, +10 more.
td_color_matching_extract_colorsColor Matching: extract coloursA
Color Matching step 0: extract the design's palette as JSON (motif and background colours with hex, pantone and percentage). Pass its task_id as extract_task_id to the suggest, match-reference and swatch-grid tools; use the hexes with td_color_matching_replace. Best for: reading a design's motif and background colours (hex, Pantone TCX... Result files are saved to disk and their paths returned.
Inputs: image (required), wait, output_dir, estimate_only, +3 more.
td_color_matching_match_referenceColor Matching: match reference coloursA
Color Matching Match from reference: recolour plans that match each reference image, from the design's extracted palette (extract_task_id or colors). Returns JSON plus mapping sets for td_color_matching_replace. Best for: matching the design's colours to the colour style of one or more reference images. Result files are saved to disk and their paths returned.
Inputs: image (required), reference_images (required), wait, colors, +6 more.
td_color_matching_regenerateColor Matching: regenerate variantsA
Color Matching Auto step 2: full-size images for the chosen cells of a swatch grid. image = the ORIGINAL design used for the grid; give grid_task_id (from td_color_matching_swatch_grid) or grid_image_url, and bg_only as in the grid run. Best for: getting clean, full-size versions of the swatch-grid variants you liked. Result files are saved to disk and their paths returned.
Inputs: image (required), variant_numbers (required), wait, bg_only, +8 more.
td_color_matching_replaceColor Matching: replace coloursA
Color Matching Replace Color: swap specific colours in the design. mappings = one set of {from, to} hex pairs; sets = several alternative sets, one output each (max 15). Get the current hexes from td_color_matching_extract_colors. Best for: swapping specific design colours for exact new values. Result files are saved to disk and their paths returned.
Inputs: image (required), sets, wait, mappings, +6 more.
td_color_matching_suggestColor Matching: suggest coloursA
Color Matching Suggest: five palette suggestions for a market and palette type, from the design's extracted palette (extract_task_id or colors). Returns JSON plus mapping sets ready for td_color_matching_replace. Best for: getting trend-aware palette ideas for a target market and palette type. Result files are saved to disk and their paths returned.
Inputs: image (required), wait, colors, market, +8 more.
td_color_matching_swatch_gridColor Matching: swatch gridA
Color Matching Auto step 1: one numbered grid image with num_options colour variants, from the design's extracted palette (extract_task_id or colors). Then call td_color_matching_regenerate with the task_id and the variant numbers you like. Best for: quickly previewing several colour variants of the design side by side (Auto mode). Result files are saved to disk and their paths returned.
Inputs: image (required), wait, colors, bg_only, +8 more.
td_color_transferColor TransferA
Color Transfer: recolour the source design with the palette of a second image. Best for: recolouring an existing design in the palette of another image, e.g. this season's trend colours or a buyer's brand colours, while keeping its structure and composition. Result files are saved to disk and their paths returned.
Inputs: image (required), palette_image (required), wait, output_dir, +4 more.
td_colour_layeringColour LayeringA
Colour Layering: separate the design into colour layers in a PSD. Advanced-Q Mode off (simple, recommended) or on (advanced); 2-30 layers or automatic. Result is a .psd file. Best for: turning flat-colour art into one editable layer per colour. Result files are saved to disk and their paths returned.
Inputs: image (required), mode, wait, n_clusters, +6 more.
td_colourwaysColourwaysA
Colourways: generate colourway options for the design (6, 12, 24 or 48), optionally keeping up to 8 of its colours. grid true also makes a buyer sheet. Best for: a buyer wants the same print in other colours, or a range needs pastel, deep and dark stories from one bestseller. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, grid, wait, +11 more.
td_design_creationDesign CreationA
Design Creation: generate design variations from the source image. Variants: prime (recommended, 1-5 outputs), new (1-5), old (1-4), reference (1-5). Best for: generating fresh design variations from a reference image without writing prompts, e.g. filling gaps in a collection, building a style-consistent range or exploring concepts quickly. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, wait, variant, +16 more.
td_design_extensionDesign ExtensionA
Design Extension: extend the design outward by a gap on each side (0-10; not all four 0). Best for: growing a small swatch or centre motif outward in the same style, colour palette and texture, e.g. a swatch photo to full fabric width or a motif out to the fabric edge. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, wait, top_gap, +9 more.
td_design_generationDesign GenerationA
Design Generation: generate a new design from the source image plus a prompt (text_to_image, recommended), re-imagine it with a texture / colour palette (image_to_image), or repaint a masked area (inpaint). The website's Mood Board is not available here. Best for: creating a new design from your image plus a written prompt. Result files are saved to disk and their paths returned.
Inputs: image (required), mode, wait, prompt, +10 more.
td_dress_to_designDress to DesignA
Dress to Design: extract the print from a garment photo as a flat design. Modes: advanced_extraction (recommended), fine_detail, inspired_pattern, all_modes (one output per mode). 1-4 outputs. Best for: getting a flat, printable 2D design from a photo of a garment, mockup or outfit (worn by a person, on a mannequin or flat-lay). Result files are saved to disk and their paths returned.
Inputs: image (required), wait, model_type, output_dir, +6 more.
td_embroidery_effectEmbroidery EffectA
Embroidery Effect: render the design as embroidery. Stitch types: bandhani, kantha, chikankari, zardozi, phulkari, aari, satin, chain, cross, running, french_knot. 13 credits. Best for: previewing a design as realistic embroidery or Indian craftwork (e.g. Zardozi, Chikankari, Phulkari) before sampling. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, wait, density, +8 more.
td_fabric_texture_removalFabric Texture RemovalA
Fabric Texture Removal: flatten the fabric weave/texture out of a photographed design. 1-4 outputs. Best for: turning a flat scan or photo of printed fabric into a clean digital design: removes the weave and surface texture while keeping the artwork and its colours, e.g. digitising an old swatch. Result files are saved to disk and their paths returned.
Inputs: image (required), wait, output_dir, num_outputs, +4 more.
td_foil_separationFoil SeparationA
Foil Separation: build the foil layer for the design. Modes outline, foiling (recommended), dot, patch; with or without a vector EPS. Best for: making a foil-ready plate from a flat, high-contrast print design, as clean outlines, distressed foiling line art, dots or accent patches. Result files are saved to disk and their paths returned.
Inputs: image (required), mode, wait, vectorize, +13 more.
td_object_layeringObject LayeringA
Object Layering: split the design into a layered PSD, one layer per object. Result is a .psd file. Best for: splitting a design into editable layers by object or motif, e.g. to move flowers apart in Photoshop instead of redrawing. Works best with distinct motifs on a contrasting background. Result files are saved to disk and their paths returned.
Inputs: image (required), wait, output_dir, estimate_only, +3 more.
td_ready_to_printReady to PrintA
Ready to Print: make a design print-ready by enlarging it 1-4x with AI detail. Best for: making a finished design print-ready: enlarging low-res buyer, archive or small-format files 1-4x with AI-filled detail and no pixelation, with your target DPI written into the file. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, wait, prompt, +9 more.
td_repeat_setRepeat SetA
Repeat Set: turn the design into a seamless repeat. Quick repeat types: block (recommended), half_brick, half_drop, mirror, step. Best for: turning a design into a seamless repeating tile for fabric, wallpaper or rotary printing, laid out as a block, half brick, half drop, mirror or step repeat. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, wait, step_axis, +11 more.
td_sketch_to_designSketch to DesignA
Sketch to Design: colour a line sketch into a finished design, optionally in a named style. Best for: turning a sketch, line art or wireframe into a full-colour, detailed design, e.g. rough brainstorm sketches into production art. Clear, simple sketches with well-defined lines work best. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, wait, style, +6 more.
td_spec_boardSpec BoardA
Spec Board: build a tech-pack style spec board from a garment photo. Give garment_type, or use Reference mode: a reference_image (an existing spec board) whose layout and sections are mimicked. Best for: turning a garment photo (saree, kurti, suit, lehenga, frock or gown) into a professional specification board for production. Result files are saved to disk and their paths returned.
Inputs: image (required), wait, output_dir, garment_type, +5 more.
td_style_transferStyle TransferA
Style Transfer: restyle the design with a library style or with the style of a second image (image_to_image). Best for: giving a finished design a new artistic surface look (e.g. Watercolor, Oil Painting) or the look of a second image, while the design layout stays intact. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, mode, wait, +7 more.
td_super_scalerSuper ScalerA
Super Scaler: AI enlargement 2x/4x/8x with detail enhancement. Best for: rescuing low-resolution, blurred, compressed or old images (72 dpi archive JPEGs, small web images) by rebuilding lost detail and enlarging up to 8x for large-format printing. Result files are saved to disk and their paths returned.
Inputs: image (required), mode, wait, output_dir, +7 more.
td_vectorizerVectorizerA
Vectorizer: convert the design to vector. ready_for photoshop (png, tiff or eps = Sharp Edged File) or illustrator (svg with curve options, pdf or eps). Best for: turning flat-colour raster art (e.g. a JPEG logo or print) into vector you can recolour, resize and edit in Illustrator, or scale up for large-format printing without quality loss. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, wait, format, +10 more.
td_watermark_removalWatermark RemovalA
Watermark Removal: remove watermarks from a design. Standard or AI Pro model. Best for: cleaning watermarks, logos, stamped branding or overlaid text off images you have the rights to modify, e.g. a residual mark on a licensed stock file or old branding on archive designs. Result files are saved to disk and their paths returned.
Inputs: image (required), dpi, wait, model, +5 more.
whoamiAccount statusARead-onlyIdempotent
Who is connected: email, organisation, credits, plan access, enabled tools. Reports a pending login if one is in progress.
Inputs: no named parameters.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 39 tools
Several overlapping clusters exist—image enhancement (td_ready_to_print, td_super_scaler, td_anti_blur), design generation (td_design_generation, td_design_creation), layering (td_object_layering, td_colour_layering, td_channeling), and colour variant tools—so an agent could misselect. The detailed 'Best for' descriptions help mitigate confusion, but boundaries are not always crisp.
Nearly all tools use snake_case, with a clear td_ prefix for design tools and no prefix for system/account tools. However, spelling is inconsistent between British 'colour' (td_colour_layering, td_colourways) and American 'color' (td_color_transfer, td_color_matching_*), and the naming mixes verb and noun styles.
39 tools is well above the 3–15 range for a well-scoped server and exceeds the 25+ threshold for 'too many'. Although each tool represents a distinct design operation, the sheer breadth creates a heavy selection burden for an agent.
The surface covers a wide range of textile design workflows: generation, enhancement, colour manipulation, layering, separation, print prep, and spec boards, plus auth, job, and documentation tools. Minor gaps exist (e.g., no explicit job cancellation or job listing), but agents can work around these via job_status and other tools.
Related MCP Connectors
Generate images from Claude, ChatGPT, Cursor and Codex with 12 top AI image models. Up to 4 images per call, SVG vectorization, background removal, upscaling and image-to-video. Remote MCP with OAuth, no API keys.
Generate images from Claude, ChatGPT, Cursor and Codex with 12 top AI image models and 10 tools: up to 4 images per call, SVG vectorization, editing and more. Remote MCP with OAuth, no API keys. Pay per image, no subscription.
Use AI models for chat, image, and video generation from Claude Code and other MCP hosts.
OCR, transcription, file extraction, and image generation for AI agents via MCP.
Related MCP Servers
- AlicenseNot gradedqualityFmaintenanceEnables AI image processing tools such as vectorization, background removal, upscaling, and logo generation from any MCP-compatible client like Claude Code or Cursor.23 npmMIT

Ads Optimiser MCP serverofficial
AlicenseAqualityBmaintenanceLets Claude Desktop, Claude Code, or any MCP client generate ad images and videos from chat, including batches from folders of local files, prompt lists, and pipelines. Sign-in uses an OAuth device flow with workspace-scoped, revocable tokens, and finished creatives can be saved back to your machine.1630 npmMIT
VectoSolve MCP Serverofficial
AlicenseNot gradedqualityCmaintenanceEnables AI image tools directly from Claude Code, Cursor, Windsurf, and other MCP-compatible clients, including vectorization, background removal, upscaling, and logo generation.23 npm1MIT- AlicenseBqualityCmaintenanceEnables Claude, Cursor, and other MCP clients to generate and re-render images on Tensor.Art using public checkpoints or private LoRAs, run ComfyUI workflows, and estimate credit costs before spending.9MIT