MCP methods reference

Everything Claude can
create for you

32 tools across generation, editing, templates, assets, and brand. Every one works from a normal chat message — no code, no copy-paste. This page is what Claude sees.

Create Edit & transform Combine Templates & layout Your library Brand themes Discovery & account
01 — Create

Make something from nothing

Ask Claude to generate images, video, and audio using the top AI models on the market. One prompt, one answer.

generate_image

Generate an image

Headline

Create a new image from a text prompt, or edit an existing one by passing an image URL. Works with Nano Banana Pro, Flux, Recraft, Ideogram, and every other image model.

Say to Claude

"Make me a product shot of a matte black espresso machine on a warm kitchen counter, golden hour light"

Say to Claude

"Take this logo and put it on a coffee cup, photorealistic"

Options (7) +
prompt string required

What to create, or edit instructions when image_url is set.

model string

Model ID. Defaults to Flux Schnell. Use list_models to browse.

image_url string

Pass a source image to switch into edit mode.

width / height number

Output dimensions in pixels.

num_images 1-4

Generate multiple variations in one call.

remove_background boolean

Auto-remove the background from the result.

extras object

Model-specific parameters (see get_model_params). Pass extras.image_urls for style/character references — the first one auto-promotes to image_url if you forgot to set it.

generate_video

Generate a video

Headline

Produce video from text, from a single start frame, from first-and-last frames, or from reference media. Returns a job to poll.

Say to Claude

"Make a 6-second vertical launch clip: camera slowly orbits a new pair of running shoes on a pedestal"

Say to Claude

"Animate this photo into a 5-second cinematic shot with subtle camera push-in"

Options (7) +
prompt string required

Description of the shot or motion.

model string

Hailuo, Kling, Veo, Seedance, and more. Defaults to Hailuo 2.3 Fast.

image_url string

Start frame for image-to-video.

last_frame_url string

End frame for first-last-frame interpolation.

duration string

Clip length. Model-dependent values.

aspect_ratio string

E.g. 9:16, 16:9, 1:1.

image_urls / video_urls / audio_urls array

Reference media for models that support it.

generate_speech

Generate speech

Headline

Turn any text into studio-quality narration with voice, speed, and emotion controls. Pass a reference audio URL to clone a voice with Chatterbox.

Say to Claude

"Narrate this script in a warm, confident female voice at a slightly slower pace"

Say to Claude

"Clone the voice in this audio sample and read this paragraph in that voice"

Options (7) +
text string required

Up to 5000 characters.

model string

ElevenLabs, MiniMax HD, Dia, Chatterbox, and more. Defaults to MiniMax HD.

voice_id string

Pick a voice from the model catalog.

audio_url string

Reference audio for voice cloning (Chatterbox).

speed 0.5-2.0

Playback speed multiplier.

emotion string

happy, sad, angry, surprised, etc (MiniMax).

language_boost string

Bias toward a non-English language.

compare_models

Compare models side-by-side

Run the same image prompt through 2-4 models in parallel and see them lined up. The fastest way to find the right model for a brief.

Say to Claude

"Compare Flux Pro, Nano Banana Pro, and Recraft V3 for a minimalist packaging shot"

Say to Claude

"Run this prompt through 3 image models so I can pick the best one"

Options (4) +
prompt string required

Shared prompt across all models.

models string[] required

Array of 2-4 model IDs.

width / height number

Shared output dimensions.

seed number

Shared seed for reproducibility.

02 — Edit & transform

Polish what you have

Upscale, trim, scale, subtitle, remove backgrounds, and pull frames out of any image or video. Each transform is its own dedicated tool.

upscale_media

Upscale image or video

One tool that upscales both images and videos. Pick the backend that fits: Aura, Clarity, or Recraft for stills; Real-ESRGAN or ByteDance for video.

Say to Claude

"Upscale this hero image 4x with Clarity, keep it sharp"

Say to Claude

"Upscale this video to 4K at 60fps with the bytedance preset"

Options (9) +
media_url string required

Source image or video URL.

type image | video required

Which kind of media.

upscale_model string

Image: aura, clarity, recraft. Video: realesrgan, bytedance.

upscale_factor 1-4

Image only. Clarity supports variable factors.

upscale_scale 1-8

Video only. Real-ESRGAN factor.

target_resolution 1080p | 2k | 4k

Video only (ByteDance).

target_fps 30 | 60

Video only (ByteDance).

enhancement_preset general | ugc | short_series | aigc | old_film

Video only (ByteDance).

prompt / creativity / resemblance mixed

Image: tune the Clarity upscaler.

trim_video

Trim a video

Cut a clip down to a specific start and end. Returns a fresh permanent URL — original is untouched.

Say to Claude

"Trim this video from 0:05 to 0:20"

Say to Claude

"Cut the first 10 seconds off this clip"

Options (4) +
video_url string required

Source video.

start_time number

Start in seconds.

end_time number

End in seconds.

duration number

Length in seconds (alternative to end_time).

scale_video

Scale or reframe a video

Resize a video to any dimensions. Stretch, pad with a background color, or crop to fit — perfect for adapting one clip to every social platform.

Say to Claude

"Scale this video to 1080x1920 for TikTok, pad with black bars"

Say to Claude

"Crop this 16:9 video to a 1:1 square"

Options (5) +
video_url string required

Source video.

width number required

Target width in pixels.

height number required

Target height in pixels.

mode stretch | pad | crop

How to fit content into the new frame.

pad_color string

Hex color for pad bars.

add_subtitles

Add subtitles to a video

Auto-transcribe and burn in animated subtitles with word-level karaoke highlighting. Full control over font, color, position, and animation.

Say to Claude

"Add bold white English subtitles with a yellow karaoke highlight, centered at the bottom"

Say to Claude

"Caption this video in Spanish, large font, 3 words per subtitle"

Options (7) +
video_url string required

Source video.

language string

Language of the audio (auto-detected if omitted).

font_name / font_size / font_weight / font_color mixed

Typography controls.

highlight_color / stroke_color / background_color string

Color styling.

position / y_offset string / number

Where the caption sits on screen.

words_per_subtitle number

How many words per subtitle line.

enable_animation boolean

Karaoke-style word highlighting.

remove_background

Remove a background

Strip the background from an image or video. Images become transparent PNGs. Videos can output VP9 (WebM with alpha) or H.264.

Say to Claude

"Remove the background from this product photo"

Say to Claude

"Cut the speaker out of this video and output as WebM with transparency"

Options (5) +
media_url string required

Source image or video.

type image | video required

Which kind of media.

refine_foreground_edges boolean

Video only. Cleaner edges, slower.

subject_is_person boolean

Video only. Optimizes for human subjects.

output_codec vp9 | h264

Video only. VP9 keeps the alpha channel.

extract_frames

Extract frames from video

Pull a single thumbnail or batch export every Nth frame as images. Great for storyboards and social thumbnails.

Say to Claude

"Grab the middle frame of this video as a thumbnail"

Say to Claude

"Export a frame every second from this clip as JPGs"

Options (6) +
video_url string required

Source clip.

mode single | batch required

One frame or many.

position first | middle | last

Single mode only.

frame_interval number

Batch mode: every Nth frame.

max_frames number

Cap the batch size.

output_format png | jpg | webp

Output image format.

03 — Combine

Stitch media together

Merge audio onto video, concatenate clips, or join audio tracks into a single deliverable.

merge_media

Merge audio, video, or both

Glue media together in three ways: lay an audio track over a video, concatenate multiple videos, or join audio clips end to end.

Say to Claude

"Take this video and this voiceover and merge them into one file"

Say to Claude

"Concatenate these 4 clips in order into a single video"

Say to Claude

"Join these three audio clips into one track"

Options (4) +
operation merge_audio_video | merge_videos | merge_audios required

Pick the combine mode.

video_url + audio_url string

For merge_audio_video.

video_urls string[]

Two or more clips to concatenate.

audio_urls string[]

Two or more audio files to join.

04 — Templates & layout

Branded graphics from HTML

Render images directly from HTML and CSS, or build reusable templates with token placeholders. Faster, cheaper, and pixel-perfect when AI generation is overkill.

render_html_image

Render HTML to a PNG

New

Turn HTML and CSS into a PNG using Satori. Perfect for OG images, social cards, and layout-driven graphics where AI generation is too slow or too expensive.

Say to Claude

"Render this HTML as a 1200x630 OG image with our brand fonts"

Say to Claude

"Make a social card with this headline on a coral background, centered"

Options (6) +
html string required

HTML/CSS to render. Inline styles only — flexbox layout.

width number

Defaults to 1200.

height number

Defaults to 630.

fonts array

Font URLs to embed.

inline_images array

Token-to-URL map for embedded images.

name / tags string

Library metadata.

create_template

Save a reusable template

New

Define an HTML template with {{token}} placeholders for headlines, images, and colors. Render it later with different values for bulk image production.

Say to Claude

"Save this HTML as a template called "launch-card" with parameters for title, subtitle, and product_image"

Say to Claude

"Create a template for our weekly newsletter header"

Options (7) +
name string required

Template identifier.

html string required

HTML with {{token}} placeholders.

width / height number required

Output dimensions.

parameters array

Each: name, type (text/image_url/color/number), default, required.

fonts array

Embedded fonts.

static_assets array

Token-to-URL map for fixed images.

description string

What the template is for.

render_template

Render a template

New

Generate an image from a saved template by passing parameter values. Override dimensions or fonts on the fly when you need variations.

Say to Claude

"Render the "launch-card" template with title "New Espresso Blend" and our hero photo"

Say to Claude

"Use the newsletter template with this week's headline"

Options (6) +
template_id uuid

Template ID. Or use template_name.

template_name string

Template name. Or use template_id.

modifications object

Token-to-value map for substitution.

width_override / height_override number

Override saved dimensions.

fonts_override array

Override saved fonts.

name / tags string

Library metadata for the output.

list_templates

List your templates

New

Discover the templates you have saved. Filter by name and optionally include the full HTML for inspection.

Say to Claude

"List all my image templates"

Say to Claude

"Find templates with "card" in the name"

Options (2) +
name_contains string

Substring match on template name.

include_html boolean

Return the full HTML body too.

05 — Your library

Organize every asset you create

Every generation is stored with a permanent URL. Search it, tag it, upload your own files, and reuse anything across conversations.

search_assets

Search your library

Find past generations and uploads by type, tag, text, or source. Newest results first.

Say to Claude

"Find every image I tagged "hero-banner" from the last month"

Say to Claude

"Search my library for videos about "launch""

Options (5) +
type image | video | audio | 3d_model

Filter by media type.

query string

Free-text match on name, description, prompt.

tags string[]

Match any of these tags.

source generated | uploaded | imported

Where it came from.

limit / offset number

Paginate results.

upload_asset

Upload from a URL

Pull a file from any public URL into your library. Returns a permanent CDN URL. For large local files, use get_upload_url instead.

Say to Claude

"Import the image at this URL and call it "hero-v2""

Say to Claude

"Save this audio file from a public URL and tag it "voiceover""

Options (4) +
type image | video | audio | 3d_model required

Media type.

content_type string required

MIME type.

url string required

Public URL to fetch from.

name / tags / description string

Library metadata.

import_media

Drag-and-drop from chat

Opens an interactive file picker right inside Claude so you (or the user) can drop media in without leaving the conversation.

Say to Claude

"Let me upload a reference image for this shoot"

Say to Claude

"Import the brand logo from my computer"

Options (1) +
context string

Optional human-readable hint shown in the picker.

update_asset

Rename, tag, describe

Organize assets so you can find them later. Set a unique name, replace the tag list, or update the description.

Say to Claude

"Tag this image as "hero" and "q2-campaign""

Say to Claude

"Rename asset abc123 to "launch-teaser-final""

Options (4) +
id uuid required

Asset ID.

name string

Unique per user.

tags string[]

Replaces the existing tag list.

description string

Free-text notes.

delete_asset

Delete an asset

Soft-delete an asset from your library. The underlying file is preserved but hidden from search.

Say to Claude

"Delete the draft image I generated earlier"

Options (1) +
id uuid required

Asset to remove.

get_upload_url

Large file upload (step 1)

Get a pre-signed URL for files bigger than 37MB. Upload directly via PUT, then call confirm_upload to activate the asset.

Say to Claude

"Give me an upload URL for a 300MB 4K video file"

Options (3) +
type image | video | audio | 3d_model required

Media type.

content_type string required

MIME type.

name / tags / description string

Library metadata.

confirm_upload

Large file upload (step 2)

Finalize the upload once the file is in storage. The asset becomes searchable and ready to use.

Say to Claude

"Confirm the upload for asset id abc123"

Options (1) +
asset_id uuid required

Returned by get_upload_url.

06 — Brand themes

Stay on-brand automatically

Save brand colors, fonts, logos, and voice guidelines once. Every generation can reference your theme so you never paste a brand kit again.

update_theme

Save a brand theme

Create or update a reusable theme with any JSON shape: colors, fonts, logos, product lines, tone rules. Shallow-merges by default.

Say to Claude

"Save a theme called "tinyworks" with these colors and this logo URL"

Say to Claude

"Update my default theme to add a new accent color"

Options (4) +
name string required

Theme identifier.

data object required

Any JSON structure — no fixed schema.

override boolean

Replace instead of shallow-merge.

set_default boolean

Make this the active theme.

get_theme

Read a theme

Fetch the full JSON for your default theme or a specific one. Useful before generating so Claude can reference your brand.

Say to Claude

"Show me my default theme"

Say to Claude

"Get the "holiday-campaign" theme"

Options (1) +
name string

Omit for the default theme.

list_themes

List all themes

See every theme you have saved, with the default marked and a summary of which top-level keys each one contains.

Say to Claude

"List all my brand themes"

delete_theme

Delete a theme

Soft-delete a theme. If it was the default, the oldest remaining theme is promoted automatically.

Say to Claude

"Delete the "legacy" theme"

Options (1) +
name string required

Theme to delete.

07 — Discovery & account

Pick models, track jobs, manage credits

The infrastructure layer. Most of this runs automatically in the background — but you can call it directly when you need control.

list_models

Browse available models

Filter the model catalog by category or search. Each entry shows the ID, capabilities, and a short description.

Say to Claude

"List all the video models"

Say to Claude

"Find the fastest image models"

Options (2) +
category image | video | speech

Filter to one media type.

query string

Free-text search (e.g. "flux", "cheap").

get_model_params

Inspect a model

Look up every parameter a model supports with types, defaults, ranges, and enum values. Needed when tuning the extras field.

Say to Claude

"Show me all the parameters for Nano Banana Pro"

Say to Claude

"What does Hailuo 2.3 accept as input?"

Options (1) +
model string required

Model ID from list_models.

get_credits_balance

Check credit balance

See how many credits you have left and optionally estimate what an upcoming generation will cost.

Say to Claude

"How many credits do I have?"

Say to Claude

"Estimate the cost of a 10-second Veo 3.1 video at 1080p"

Options (3) +
estimate_model string

Model to price-check.

estimate_type image | video | speech | 3d_model | video_edit

Operation category.

estimate_params object

Width, height, duration — whatever affects cost.

get_credits_link

Top up credits

Get a Polar checkout link to buy more credits. $10 for 1000 credits, one-time purchase, no subscription.

Say to Claude

"Give me a link to buy more credits"

Options (1) +
product string

Pack size. Defaults to "1000".

check_job

Check a generation job

Poll an async job (video generation, text-to-image) until it finishes. Returns the final URL once ready.

Say to Claude

"Is my video generation job done yet?"

Say to Claude

"Check on job id abc-123"

Options (1) +
job_id uuid required

ID returned by generate_image or generate_video.

load_image

Inspect an image inline

Fetch an image URL as base64 so Claude can see it and iterate. Used when you explicitly want Claude to review a visual.

Say to Claude

"Load this image so you can review it and suggest improvements"

Options (1) +
url string required

Image URL. Files must be under 800KB.

Connect it in under a minute

One MCP URL. Every tool on this page becomes available inside Claude the moment you connect.

How to connect See pricing