budgetpixel.com MCP server
com.budgetpixel/mcpbudgetpixel.comstreamable-httpv1.0.0Scanned
Generate AI images, video, music, and sound effects, and upscale them, from any MCP client.
31 tools: 14 can delete or cancel things, 6 change things and 11 only read.
- Reads private data · 1
- Reads text others write · 17
- Can send data out · 1
- 4 findings
Tools listed without sign-in. Remote posture
With Anlyon
14 tools here can delete or cancel things. Govern them.
- Anlyon lists the tools and adopts none.
- Each tool you adopt gets approvals and receipts.
- Your agent never holds the key.
A tool your agent calls directly is not routed through Anlyon and is not governed by it. Anlyon Radar only lists and labels, and these labels are the same either way.
What the budgetpixel.com MCP server can do
The budgetpixel.com MCP server is published by budgetpixel.com, a domain the MCP Registry verified. It has 31 tools. It can delete or cancel things (14 tools: generate_image, generate_video, finalize_video_draft, generate_music and 10 more) and create or change data (6 tools: save_to_assets, upload_media, recommend_influencers, add_wardrobe_item and 2 more). 11 tools only read.
1 tool reads private data (get_character); 17 tools bring in text other people can write (generate_image, generate_video, generate_music and 14 more); 1 tool can send data out of your system (create_social_post). On its own it has all three parts of a prompt-injection exfiltration path: private data, text others write, and a way out.
The fixed checks matched 4 findings (annotation-conflict). Each one comes with the text it matched.
Its remote endpoint lists its tools without sign-in, and its tools changed 2 times since Anlyon Radar first saw them.
Last scanned Nov 10, 2026, version 1.0.0.
Tools
Grouped by what each tool does, most consequential first. Open one for its reasons and full description; descriptions are the publisher's own words.
31 tools
Deletes or cancels 14Claude Code: deny
create_social_postPublish a post to the user's PUBLIC BudgetPixel community feed — visible to everyone and labeled "via MCP".Reads text others writeCan send data outNo findings
Why it is in this group
- The name "create_social_post" says "create": a word for a change.
- The description's first verb is "publish": a word for a change that another call does not undo.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
- outboundThe name "create_social_post" says "create" and "post": it writes something other people read.
Description, as the server wrote it Publish a post to the user's PUBLIC BudgetPixel community feed — visible to everyone and labeled "via MCP". ONLY use this when the user EXPLICITLY asks to post/share/publish; never post proactively. Costs 10 credits per post (anti-spam, charged on publish — same as the developer API) with an hourly posting cap. media: 1–6 images, or 1 video + up to 3 images, or exactly 1 audio track — pass the URLs you got from generations (check_generation_status links) or upload_media, or any public URL/base64. Images can carry optional model/prompt generation info. Audio posts accept song_name, genre (a fixed list — see the enum), lyrics, AND cover_image (cover art goes in its own field, never as a second media item). To enter the post in one of the user's clan themes, pass theme_id (call list_my_themes to find it). Posts are All Rights Reserved by default; pass license ("cc-by" or "cc0") ONLY when the user explicitly asks for a Creative Commons license — it is permanent and cannot be changed after publishing. Returns the public post URL — share it with the user. Posts are moderated after publish (NSFW/spam rating affects feed visibility; the charge is not refunded for hidden posts).
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
design_poseUse this when the user wants a body pose the library doesn't have (try search_poses first).Reads text others writeNo findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it Use this when the user wants a body pose the library doesn't have (try search_poses first). Describe the pose in words, head to toe (up to 600 characters), or pass based_on with a library pose's handle and say what to change ("the left arm raised higher"). If library poses already look like the request, they come back first and no design starts: use one, or call again with confirm: true to design anyway. A design takes about 3-6 minutes: poll check_pose_design with the returned design_id every 20-30 seconds; when it is done, use the new @handle like any pose ("@barry in @<new_handle>"). 100 credits, charged only when the pose is made: a refused or failed design is free. Paid plans, up to 20 designs a day. Every pose must be one a real body can do: impossible poses, contortion, sex positions, minors, real people and violence are refused with the reason and, when there is one, a doable alternative. The new pose joins the shared library for everyone; there are no private poses.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
finalize_video_draftUse this when the user likes a finished seedance-2.5 DRAFT (480p) and wants it in 1080p: it renders the same video (same motion, framing and sound, sharper) as a new 1080p job.No findings
Why it is in this group
- The name "finalize_video_draft" says "finalize": a word for a change that another call does not undo.
- The server marks it destructive (destructiveHint: true).
Description, as the server wrote it Use this when the user likes a finished seedance-2.5 DRAFT (480p) and wants it in 1080p: it renders the same video (same motion, framing and sound, sharper) as a new 1080p job. Takes only the draft's job_id; everything else comes from the draft. IMPORTANT — cost: 750 credits per second of the draft's length (plus 375 per second of any reference video the draft used), e.g. a 5s draft's final ≈ 3,750 credits; check_generation_status on the draft shows the exact standard price. Confirm with the user first. Works for 7 days after the draft was made. Charged on success only. Returns a new job_id — poll check_generation_status every 15–30 seconds, then share the finished video link.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: false
- destructiveHint: true
generate_3d_modelTurn one picture into a textured 3D model (image-to-3D).Reads text others writeNo findings
Why it is in this group
- The name "generate_3d_model" says "generate": a word for a change.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it Turn one picture into a textured 3D model (image-to-3D). Always returns a GLB; meshy-7.1 also offers FBX, USDZ (iPhone AR) and STL (3D printing). Pick the model by job: tripo-h3.1 (default) is the most faithful reconstruction — including cartoons and stylized art — with PBR materials, ~4 min, 400 credits; meshy-7.1 makes lighter ~30K-triangle models in more formats and supports rigging (with a walk and a run) + animation for humanoid characters, the move chosen with animation_move (a dance, a wave, a cheer and hundreds more), ~2 min, 1600 credits. Best input is one subject on a plain background. Submits an async job and returns a job_id — 3D models take 2-5 minutes, so poll check_generation_status with this job_id and type "3d" every 15-30 seconds; when done it returns the thumbnail inline plus download links (GLB/FBX/USDZ/STL). Files are kept about 24 hours, so download them promptly (save_to_assets does not take 3D models). Charged on success only.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
generate_depth_mapUse this to turn a video into a depth map video (Video Depth Anything, on BudgetPixel's own GPUs): a steady, flicker-free video where near is white and far is black, at the input's…Reads text others writeNo findings
Why it is in this group
- The name "generate_depth_map" says "generate": a word for a change.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it Use this to turn a video into a depth map video (Video Depth Anything, on BudgetPixel's own GPUs): a steady, flicker-free video where near is white and far is black, at the input's size (up to 1080p) and frame rate (up to 30 fps). Use it as a depth / structure / motion reference or control video for video models — grayscale (the default colormap) is what they take as a depth reference; turbo, inferno, magma and viridis colour it for viewing only. Priced per second of the INPUT video, rounded up to whole seconds, 5-second minimum, by model size: small 20 / base 30 / large 40 credits per second (a 5-second clip or shorter costs 100 / 150 / 200, a 10-second clip 200 / 300 / 400). The resolution does not change the price. small is fastest; base adds edge detail; large draws the finest detail (hair, thin poles, distant people) and is the steadiest. Input: a video URL (pass one from upload_media for local files), up to 30 seconds and 50MB. Options: size (default small), colormap (default grayscale), resolution (default auto = the input's size, never upscaled), keep_audio (copy the input's audio), side_by_side (input on the left, depth map on the right), face_mesh (no extra charge: draws each face's mesh so a video model keeps lip, eye and head movement; set it when people talk or emote), max_faces (1-4, default 4), include_face_landmarks (no extra charge: also returns a face landmarks JSON link). The price is held when the job starts and charged on success only; a failed job is never charged. Jobs take about 1-2 minutes: the tool returns a job_id — poll check_generation_status with that job_id and type "depth" every 15-30 seconds until it returns the finished depth video, then tell the user. The depth video is also saved to the user's BudgetPixel library.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
generate_imageUse this to create or edit an AI image.Reads text others writeNo findings
Why it is in this group
- The name "generate_image" says "generate": a word for a change.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it Use this to create or edit an AI image. Text-to-image from a prompt, OR image-to-image / editing / multi-reference by passing input_images (URLs, data URIs, or base64) — e.g. edit or restyle an image the user just generated by passing its URL. Input support is per-model: some take 1 image, some several, some none (check each model's max_input_images via list_models). Submits an async job and returns a job_id — then call check_generation_status to poll (usually ready in seconds); when complete it returns the finished image inline for you to display. Each call starts a new job that spends credits from the user's BudgetPixel balance (failed jobs aren't charged). Default model: seedream-5.0-pro; call list_models to see every available model with prices and input limits. Write @handle in the prompt to cast a character (list_characters), e.g. a free BudgetPixel AI influencer like @barry; an item's @handle (search_wardrobe) dresses them: "@barry wearing @camel_coat_set"; a pose's @handle (search_poses) poses them: "@barry in @sitting_knee_up_pose".
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
generate_musicUse this to create an AI music track (full songs with vocals, or instrumentals).Reads text others writeNo findings
Why it is in this group
- The name "generate_music" says "generate": a word for a change.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it Use this to create an AI music track (full songs with vocals, or instrumentals). Three models: music-3.0 (default, 200 credits — songs from a prompt and/or lyrics, lyrics optimizer on by default so a prompt alone works), mureka-v9 (60 credits — sings the user's EXACT lyrics verbatim, or instrumental/BGM), and lyria-3 (100 credits — high-fidelity from a rich prompt, can condition on up to 10 reference images), and sonilo-music (instrumental/background music of an EXACT length via duration, 5–360 s — priced PER SECOND at 4 credits/s with a 10-second minimum, no vocals). VIDEO-TO-MUSIC: pass a `video` URL (upload_media first for local files) and sonilo-video-music composes a licensed soundtrack for the clip's content, pacing and mood — priced per second of the video's measured length (15 credits/s, 10 s minimum, ≤360 s). Flat price per track (per second for the Sonilo models), charged on success only; each call starts a new paid job. Submits an async job and returns a numeric job_id — music takes a minute or three: poll check_generation_status with the job_id AND type "music" every 15–30 seconds, then share the finished track's listen/download link with the user (chat can't embed audio players).
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
generate_sound_effectUse this to create sound effects (NOT music).Reads text others writeNo findings
Why it is in this group
- The name "generate_sound_effect" says "generate": a word for a change.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it Use this to create sound effects (NOT music). TEXT mode: describe the sound — Foley, ambience, UI sounds, whooshes/transitions, impacts, action sounds — with sonilo-sfx: 1–180 seconds, mp3 or wav, PRICED PER SECOND of the requested duration (5 credits/s, 3-second minimum → 5 s = 25, 10 s = 50, 30 s = 150 credits). VIDEO mode: pass a video URL (upload local files with upload_media first) to get frame-accurate effects synced to the clip's visible events with sonilo-video-sfx — billed per second of the VIDEO's measured length (15 credits/s, 3-second minimum, up to 360 s) and returns both the SFX track and the video with the effects mixed in. Each call starts a new paid job, charged on success only. Keep text clips as short as the sound needs and confirm cost for long videos. Submits an async job and returns a numeric job_id — text clips finish within ~15 seconds, video jobs can take a few minutes: poll check_generation_status with the job_id AND type "sound_effect" every 5–15 seconds, then share the finished link(s) with the user (chat can't embed players). For songs or background music use generate_music instead.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
generate_videoUse this to create or edit an AI video.Reads text others writeNo findings
Why it is in this group
- The name "generate_video" says "generate": a word for a change.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it Use this to create or edit an AI video. Text-to-video from a prompt; image-to-video by passing image (start frame, + optional end_image); reference-to-video by passing reference_images; or video-to-video editing by passing video (an input video URL — upload local files with upload_media first). IMPORTANT — cost: video is priced per SECOND and by resolution (default model seedance-2.5: 480p=150, 720p=330, 1080p=750 credits/sec, plus 80/160/375 per second of reference-video INPUT), so e.g. 5s at 720p ≈ 1,650 credits and 5s at 1080p ≈ 3,750; confirm with the user before expensive combos (long durations or many reference clips). Each call starts a new paid job, charged on success only. Submits an async job and returns a job_id (uuid) — video takes MINUTES: poll check_generation_status every 15–30 seconds until it returns the finished video link, then share that link with the user. Default model: seedance-2.5; call list_models for available video models and prices. Write @handle in the prompt to cast a character (list_characters), e.g. a free BudgetPixel AI influencer like @barry; an item's @handle (search_wardrobe) dresses them: "@barry wearing @camel_coat_set"; a pose's @handle (search_poses) poses them: "@barry in @sitting_knee_up_pose".
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
make_pose_viewUse this when a pose lacks the camera angle the user wants (get_pose lists the views a pose has).Reads text others write1 finding
Why it is in this group
- The name "make_pose_view" says "view": a word for reading.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Findings
- Warn
The server marks "make_pose_view" destructive (destructiveHint: true), and its name says "view", a word for reading.
What this check looks forMatched text
annotations: {"readOnlyHint":false,"destructiveHint":true,"openWorldHint":true}; The name "make_pose_view" says "view": a word for reading.
Description, as the server wrote it Use this when a pose lacks the camera angle the user wants (get_pose lists the views a pose has). It makes a view of one of the pose's mannequins from another place on the Camera Angles globe: the mannequin's front picture is turned there, a check confirms it is the same pose and where the camera landed, and the view joins the shared pose library for EVERYONE, with its own mention (there are no private views). figure: woman or man, the one matching the character's gender (a mention sends that one; woman when unknown). azimuth, where the camera stands, in steps of 45: 0 front, 45 front right, 90 right side (the camera moved to the viewer's right, so the figure's LEFT side shows), 135 back right, 180 back, 225 back left, 270 left side, 315 front left (the pose's own three-quarter picture). elevation: 0 eye level, 30 from above, 60 from high above. A place that already has a view answers at once, free. Otherwise a job starts: it usually takes 30-90 seconds, so poll check_pose_view with the returned job_id every 10-15 seconds. 40 credits, charged only when the view lands where asked (a second try is free); a try that lands on another empty place is kept there for everyone, free. If someone is already making that view, you join their job for free. Up to 50 new views a day. Once done, write the view's mention after a character: "@barry in @seiza_pose.back" (.front_right, .right, .back_right, .back, .back_left, .left, .three_quarter, .above = front from above, .top = front from high above; other places a code like .az135_el30).
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
search_stock_audioUse this when the user needs an existing sound effect or a piece of background music — for a video, podcast, stream, game, app or presentation — rather than a newly generated one.1 finding
Why it is in this group
- The name "search_stock_audio" says "search": a word for reading.
- The server marks it destructive (destructiveHint: true).
Findings
- Warn
The server marks "search_stock_audio" destructive (destructiveHint: true), and its name says "search", a word for reading.
What this check looks forMatched text
annotations: {"readOnlyHint":false,"destructiveHint":true,"openWorldHint":false}; The name "search_stock_audio" says "search": a word for reading.
Description, as the server wrote it Use this when the user needs an existing sound effect or a piece of background music — for a video, podcast, stream, game, app or presentation — rather than a newly generated one. Searches BudgetPixel's free audio library, licensed CC BY 4.0 for commercial use, every item listened to and scored before listing. kind "sound_effect": whooshes, impacts, UI sounds, Foley, ambiences, animals, vehicles (a fraction of a second to about 6 s). kind "music": instrumental background tracks such as lofi, corporate, cinematic, acoustic (up to about a minute). Costs 10 credits per search, charged even when nothing matches. Try it before generate_sound_effect or generate_music: a search is cheaper and instant. Write the query as a few plain words naming the sound ("door creak", "soft notification chime", "lofi study beat"), not a generation prompt: every word counts, and a word the library does not have counts as a miss. It never generates: if nothing fits, say so. LICENCE: wherever a result is used, include its credit line exactly as returned (the `attribution` text; `attribution_html` for web pages).
Annotations the server declares
- readOnlyHint: false
- openWorldHint: false
- destructiveHint: true
search_stock_imagesUse this when the user needs an existing stock photo, illustration, background, wallpaper or texture — for a website, slide, blog post, app, video or design — rather than a newly g…Reads text others write1 finding
Why it is in this group
- The name "search_stock_images" says "search": a word for reading.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe description's first sentence says "website": it reads web pages, which anyone can write.
Findings
- Warn
The server marks "search_stock_images" destructive (destructiveHint: true), and its name says "search", a word for reading.
What this check looks forMatched text
annotations: {"readOnlyHint":false,"destructiveHint":true,"openWorldHint":false}; The name "search_stock_images" says "search": a word for reading.
Description, as the server wrote it Use this when the user needs an existing stock photo, illustration, background, wallpaper or texture — for a website, slide, blog post, app, video or design — rather than a newly generated image. Searches BudgetPixel's stock image library: AI-generated images, each reviewed before listing (no real people, logos or text), licensed CC BY 4.0 for commercial use. Set transparent to true for cut-outs only: PNG files with a transparent background, for stickers, overlays and compositing. Costs 10 credits per search, charged even when nothing matches, so write one specific query rather than many vague ones. It never generates: if nothing fits, say so. LICENCE: wherever an image is used, include its credit line exactly as returned (the `attribution` text; `attribution_html` for web pages). Download from the returned URLs rather than hotlinking them in production.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: false
- destructiveHint: true
upscale_imageUse this to upscale an image to a higher resolution.Reads text others writeNo findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it Use this to upscale an image to a higher resolution. Default model p-image-upscale (Pruna AI) returns results in seconds and is priced by the target resolution: 4 MP = 10, 8 MP = 15, 16 MP = 30, 32 MP = 60 credits. Alternative clarity-upscaler does creative detail enhancement (steer with prompt/creativity) at a flat 30 credits and takes ~60 seconds. Pass the image as a URL (a generated image's URL, or upload_media for local files), a data URI, or base64. Charged on success only. The tool waits for the result and returns the upscaled image's URL — share it with the user. Unlike generations, the upscaled image is also saved to the user's BudgetPixel library.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
upscale_videoUse this to upscale a video, with one of two models.Reads text others writeNo findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it Use this to upscale a video, with one of two models. Topaz (model topazlabs-video-upscale, the default): 1080p or 4K at 30 or 60 fps, priced per second of the INPUT video: 1080p = 15 (30fps) / 30 (60fps) credits/s, 4K = 60 / 120 credits/s, so a 10s clip to 1080p30 costs 150 credits; its result is an H.265 (HEVC) 10-bit MP4, which some browsers cannot play (black frame with audio) — tell the user. FLUX 3 Video Upscale (model flux-3-video-upscale): 1.5x to 3x the input frame (input up to 2560x1440), precise (keeps faces and products) or creative (adds fine detail), keeps the frame rate and audio, H.264 output; priced per megapixel-second of OUTPUT: 95 credits precise / 135 creative, so one second of 1080p output is 188 / 267 credits and of 4K 752 / 1068 — it costs far more than Topaz, so state the price and confirm with the user first. Input up to 20 seconds (pass a URL; upload_media for local files). The price is held when the job starts and charged on success only. Jobs take a few MINUTES: the tool returns a job_id — call upscale_video again with just job_id every 30-60 seconds until it returns the finished video URL, then share it. The upscaled video is also saved to the user's BudgetPixel library.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: true
Changes things 6Claude Code: ask
add_wardrobe_itemUse this to add clothes to the user's wardrobe so a character can wear them.Reads text others writeNo findings
Why it is in this group
- The name "add_wardrobe_item" says "add": a word for a change.
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it Use this to add clothes to the user's wardrobe so a character can wear them. Give a picture (image_url: any picture's own address, a shop's product page, or a URL from upload_media; or image_base64) or a description. A product page is read for its product picture, but many shops (Amazon, Zara, H&M, eBay ...) don't allow automatic reading or show no product picture: the tool then says the page can't be read and what to do. Ask the user for the picture's own address (right-click the picture, Copy image address) or the picture itself, or, if you can browse, open the page yourself and pass the product picture's address. mode: photo (default for a picture: someone wears it and the clothes are taken out onto an invisible mannequin; charged at the image model's usual price, like generate_image), clean (the picture shows only the clothes: used as it is, free) or text (designed from the description, the model's usual price). note says what to take when a photo shows more ("only the leather jacket"). The picture takes about a minute: call get_wardrobe_item with the returned id until it is ready, then write its @handle next to a character in generate_image or generate_video.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: true
- destructiveHint: false
check_pose_designUse this after design_pose to follow the design: pass its design_id every 20-30 seconds until the status is done, refused or failed (a design takes about 3-6 minutes).No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it Use this after design_pose to follow the design: pass its design_id every 20-30 seconds until the status is done, refused or failed (a design takes about 3-6 minutes). Done: the new pose's @handle, description and pictures; write "@<character> in @<new_handle>" in generate_image or generate_video. Refused: why, and a doable alternative to design instead; nothing is charged. Free.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
recommend_influencersUse this when the user wants an AI influencer or a character for a product, brand, campaign, ad or post and hasn't picked one.No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it Use this when the user wants an AI influencer or a character for a product, brand, campaign, ad or post and hasn't picked one. Pass the brief (what it is, for whom, where it will be posted) and get the BudgetPixel AI influencers who fit it best, each with why, a content idea and a ready-to-use prompt that starts with their @handle (pass it to generate_image or generate_video). include_mine also considers the user's own characters. Free.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
save_to_assetsKeep a generation permanently.No findings
Why it is in this group
- The name "save_to_assets" says "save": a word for a change.
Description, as the server wrote it Keep a generation permanently. Outputs made through this connector expire about 24 hours after generation and are NOT stored in the user's BudgetPixel assets by default. Call this when the user wants to keep, reuse later, or find a result in their gallery / workshops — it copies the output into their assets (this counts against their storage like a website generation). Pass the job_id (and type for video / music / sound_effect). Safe to call twice: an already-saved job just reports already_saved.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: false
- destructiveHint: false
- idempotentHint: true
search_posesUse this to put a character in a specific body pose (sitting with one knee up, hand on hip, a yoga stance, mid-jump ...).1 finding
Why it is in this group
- The name "search_poses" says "search": a word for reading.
- The description's first verb is "put": a word for a change.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Findings
- Warn
The server marks "search_poses" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"openWorldHint":false}; The name "search_poses" says "search": a word for reading.
Description, as the server wrote it Use this to put a character in a specific body pose (sitting with one knee up, hand on hip, a yoga stance, mid-jump ...). It searches the shared BudgetPixel pose library: every pose is drawn on two faceless grey mannequins, a woman and a man, and every pose is one a real body can do. Then write "@<character> in @<pose_handle>" in the prompt of generate_image or generate_video, e.g. "@barry in @sitting_knee_up_pose on a beach at sunset": the woman or man mannequin is picked automatically from the character's gender, and the character takes the position while keeping their own face, body and clothes. Add an angle to the handle for another view, e.g. @sitting_knee_up_pose.three_quarter (front by default; get_pose lists the angles a pose has). query searches by words ("knee up", "yoga"); shelf narrows to one kind of pose. When nothing fits, design_pose makes a new one. Free.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
upload_mediaUpload a local image, video, or audio file to BudgetPixel and get back a short-lived URL (valid ~24h) to use as a generation input: pass it as an input_images value on generate_ima…No findings
Why it is in this group
- The name "upload_media" says "upload": a word for a change.
Description, as the server wrote it Upload a local image, video, or audio file to BudgetPixel and get back a short-lived URL (valid ~24h) to use as a generation input: pass it as an input_images value on generate_image, or as image / end_image / reference_images / video / audio on generate_video. Use this when the user has a LOCAL file: read the file and pass its base64 as 'file'. Not charged; max 50MB. Video/audio inputs REQUIRE this (generate_video takes them by URL only). If you already have a public URL, pass it straight to the generate tool instead.
Annotations the server declares
- readOnlyHint: false
- openWorldHint: false
- destructiveHint: false
Only reads 11Show 11 read-only tools
check_generation_statusUse this after calling generate_image, generate_video, generate_music, generate_sound_effect, generate_3d_model, or generate_depth_map to poll for results.No findings
Why it is in this group
- The name "check_generation_status" says "status": a word for reading.
Description, as the server wrote it Use this after calling generate_image, generate_video, generate_music, generate_sound_effect, generate_3d_model, or generate_depth_map to poll for results. Pass a job_id to check a specific job — numeric = image, uuid = video, and for MUSIC / SOUND-EFFECT / 3D-MODEL / DEPTH-MAP jobs also pass type "music" / "sound_effect" / "3d" / "depth" (3D jobs are uuids too, like video) — with no job_id it lists jobs still running that were started on the BudgetPixel website. Completed IMAGE jobs return the finished image(s) inline (as image content) plus a full-resolution link — display the image AND share the link (some chat UIs don't show inline images). Completed VIDEO, MUSIC and SOUND-EFFECT jobs return a watch/listen link — always share it with the user (chat UIs can't embed players). Completed 3D-MODEL jobs return the thumbnail inline plus download links (GLB/FBX/USDZ/STL) and a viewer link — share the links and tell the user downloads expire in about 24 hours. Completed DEPTH-MAP jobs return the depth video, which is saved in the user's BudgetPixel library. Poll every few seconds for images and sound effects (~15 s), every 15–30 seconds for videos, music, 3D models and depth maps (they take minutes). Call repeatedly until status is 'completed' or 'failed'.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
check_pose_viewUse this after make_pose_view to follow the view: pass its job_id every 10-15 seconds until the status is done, missed or failed (a view usually takes 30-90 seconds).No findings
Why it is in this group
- The name "check_pose_view" says "view": a word for reading.
Description, as the server wrote it Use this after make_pose_view to follow the view: pass its job_id every 10-15 seconds until the status is done, missed or failed (a view usually takes 30-90 seconds). Done: the view's picture, now in the shared library for everyone, and the mention that asks for it; write "@<character> in @<pose_handle>.<view>" in generate_image or generate_video. Missed (two tries did not come out right) and failed cost nothing; a try that landed on another empty place is kept there for everyone, free, and its mention is given too. Free.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
get_characterUse this to learn one character before casting it or using it elsewhere: its look word for word, personality, speaking style, backstory, directing notes, signature details and cons…Reads private dataReads text others writeNo findings
Why it is in this group
- The name "get_character" says "get": a word for reading.
What it can reach
- private-dataThe description's first sentence says "note": it reads notes.
- untrusted-inputThe description's first sentence says "mention": it reads comments, which other people can write.
Description, as the server wrote it Use this to learn one character before casting it or using it elsewhere: its look word for word, personality, speaking style, backstory, directing notes, signature details and consistency rules, every picture with a download link (full size for the BudgetPixel influencers and the user's own characters, so they can be used with other tools too), the models a mention works on, and the terms for using an influencer. Use the profile to write prompts, captions and scripts in the character's own voice. Free.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
get_credit_balanceUse this when the user asks about their remaining credits, subscription plan, or usage.No findings
Why it is in this group
- The name "get_credit_balance" says "get": a word for reading.
Description, as the server wrote it Use this when the user asks about their remaining credits, subscription plan, or usage. Returns monthly credit balance, extra credits, total available, and plan name.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
get_generation_historyGet the user's recent generations made through this connector or the developer API — images by default, or pass type "video", "music", or "sound_effect".No findings
Why it is in this group
- The name "get_generation_history" says "get": a word for reading.
Description, as the server wrote it Get the user's recent generations made through this connector or the developer API — images by default, or pass type "video", "music", or "sound_effect". Outputs are kept for about 24 hours (they are never saved to the user's asset storage unless save_to_assets is called), so rows older than that show as expired with no link. Returns prompts, models, status, and result links: image URLs (reusable as input_images on generate_image to edit/restyle one — e.g. "make the third one wintery"), video watch links, or music / sound-effect listen links (also reusable as create_social_post media). Every row's job_id works with check_generation_status (music and sound effects also need their type) and with save_to_assets. Generations the user made on the BudgetPixel website are in their gallery, not here.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
get_poseUse this to read one pose before using it: its description word for word (the text a mention sends with the mannequin picture), shelf, difficulty, what carries the body, props, and…Reads text others writeNo findings
Why it is in this group
- The name "get_pose" says "get": a word for reading.
What it can reach
- untrusted-inputThe description's first sentence says "read" and "mention": it reads comments, which other people can write.
Description, as the server wrote it Use this to read one pose before using it: its description word for word (the text a mention sends with the mannequin picture), shelf, difficulty, what carries the body, props, and every mannequin picture by figure (woman, man) and place on the Camera Angles globe: the pose's own pictures and the views members made, each with its place in words and numbers ("Back left · from above", azimuth 225, elevation 30) and the mention that asks for that view ("@barry in @seiza_pose.back"). The woman or man mannequin is picked automatically from the character's gender. When the angle the user wants is missing, make_pose_view makes it. Free.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
get_wardrobe_itemUse this to read one wardrobe item before using it: its description word for word (the text a mention sends with the picture), colours, tags, a download link to its picture and how…Reads text others writeNo findings
Why it is in this group
- The name "get_wardrobe_item" says "get": a word for reading.
What it can reach
- untrusted-inputThe description's first sentence says "read" and "mention": it reads comments, which other people can write.
Description, as the server wrote it Use this to read one wardrobe item before using it: its description word for word (the text a mention sends with the picture), colours, tags, a download link to its picture and how to use it. Pass the id that add_wardrobe_item returned to see whether a new outfit is ready yet (and its final @handle). Free.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
list_charactersUse this to see the characters the user can cast in images and videos: the free BudgetPixel AI influencers (memorable personas with a consistent face, like @barry, @rosa or @fifi, …No findings
Why it is in this group
- The name "list_characters" says "list": a word for reading.
Description, as the server wrote it Use this to see the characters the user can cast in images and videos: the free BudgetPixel AI influencers (memorable personas with a consistent face, like @barry, @rosa or @fifi, free for every user) plus the user's own and collected characters. To cast one, write its @handle in the prompt of generate_image or generate_video (e.g. "@barry orders coffee like he owns the cafe") on a model that takes characters. q filters by words (e.g. "fitness"); source narrows to official, mine or collected. For one character's full profile and pictures call get_character; to pick influencers for a brief call recommend_influencers. Free.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
list_modelsList the AI image, video, music, sound-effect, and 3D models available on BudgetPixel, plus the depth-map tool's sizes, with standard credit prices and capabilities.No findings
Why it is in this group
- The name "list_models" says "list": a word for reading.
Description, as the server wrote it List the AI image, video, music, sound-effect, and 3D models available on BudgetPixel, plus the depth-map tool's sizes, with standard credit prices and capabilities. Featured models come first with a one-line role hint (when to pick each). Image models that charge by output size list every size's price (the size you pass as resolution sets the price). Video models are priced per SECOND by resolution; music models are flat per track; sound effects are per second with a 3-second minimum; 3D models are priced per job (base price plus named options); depth maps (generate_depth_map) are priced per second of the input video by model size (small / base / large), 5-second minimum. Prices shown are standard rates in credits. Video resolutions delivered as H.265 (HEVC) instead of H.264 are marked, because some browsers cannot play H.265.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
list_my_themesList the live clan themes the user can post to — themes of the clans they own, moderate, or have joined (each with the user's role, the theme's brief, and when entries close).Reads text others writeNo findings
Why it is in this group
- The name "list_my_themes" says "list": a word for reading.
What it can reach
- untrusted-inputThe description's first sentence says "list" and "post": it reads posts, which other people can write.
Description, as the server wrote it List the live clan themes the user can post to — themes of the clans they own, moderate, or have joined (each with the user's role, the theme's brief, and when entries close). Use it when the user wants to post into a theme / enter a theme / share to their clan, then pass the chosen theme's id as create_social_post's theme_id. Free and read-only; themes that haven't started or have ended are not listed.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
search_wardrobeUse this to find clothes to put on a character: the user's own wardrobe items, the ones they saved and the free BudgetPixel library (dresses, suits, sportswear, costumes ...).No findings
Why it is in this group
- The name "search_wardrobe" says "search": a word for reading.
Description, as the server wrote it Use this to find clothes to put on a character: the user's own wardrobe items, the ones they saved and the free BudgetPixel library (dresses, suits, sportswear, costumes ...). To dress a character, write both @handles in the prompt of generate_image or generate_video, e.g. "@barry wearing @camel_coat_set at a gala": the item's picture goes with it and the person wears it. q filters by words ("red dress", "winter"); category by shelf (work, party, sport, swim, underwear ...); source narrows to mine, saved or house. fit is who the clothes are cut for: pass the character's gender (from list_characters or get_character: female or male) to get what suits them, unisex clothes included. Every item says its fit. Free.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: false
- destructiveHint: false
This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.
Exfil surface
Which tools read private data, which read text someone else can write, and which can send data out. One of each is a path a prompt injection can use.
17 paths in this server alone from private data, through text someone else wrote, to a way out. Denying 1 outbound tool breaks every one.
- Reads private data1
- Reads files, mail, messages, databases, secrets, calendars or repositories.
- Reads text others write17
- Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
- Can send data out1
- Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.
{
"permissions": {
"deny": [
"mcp__budgetpixel__create_social_post"
]
}
}Other servers in the same agent add their own tools to these paths. Audit your whole setup to see them together.
Remote posture
What anyone on the internet sees at the remote endpoint: an anonymous initialize and tools/list, and the public OAuth metadata.
Looked at Nov 20, 2026, 18:35 UTC, unchanged since Nov 6, 2026. No credentials were sent and no tool was called. What is looked at.
- Reachable
- yes
- HTTP status of an anonymous initialize
- 200
- tools/list served without credentials
- yes
- Tools that change things, listed without credentials
- 20
- Protocol version
- 2025-06-18
- WWW-Authenticate on 401
- not observed
- resource_metadata in WWW-Authenticate
- not observed
- RFC 9728 protected resource metadata
- https://mcp.budgetpixel.com/.well-known/oauth-protected-resource
- Authorization servers
- https://mcp.budgetpixel.com
- Legacy HTTP+SSE transport
- no
- Dynamic client registration
- yes
- Client ID metadata documents
- no
- HighThe whole server
The endpoint served tools/list without credentials, and the list has 20 tools that change things (14 destructive, 6 write). Whether a call also works without credentials was not tested: tools are never called.
What this check looks forMatched text
https://mcp.budgetpixel.com/mcp: tools/list returned 31 tools to an anonymous client, including generate_image (destructive), generate_video (destructive), finalize_video_draft (destructive), generate_music (destructive), generate_sound_effect (destructive) and 15 more
Posture changes
- Nov 6, 2026
- Tools that change things, listed without credentials: 14 to 20
Findings
Matches of fixed checks, with the text they matched.
The server marks "search_stock_images" destructive (destructiveHint: true), and its name says "search", a word for reading.
What this check looks forMatched text
annotations: {"readOnlyHint":false,"destructiveHint":true,"openWorldHint":false}; The name "search_stock_images" says "search": a word for reading.
The server marks "search_stock_audio" destructive (destructiveHint: true), and its name says "search", a word for reading.
What this check looks forMatched text
annotations: {"readOnlyHint":false,"destructiveHint":true,"openWorldHint":false}; The name "search_stock_audio" says "search": a word for reading.
- Warnsearch_poses
The server marks "search_poses" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"openWorldHint":false}; The name "search_poses" says "search": a word for reading.
- Warnmake_pose_view
The server marks "make_pose_view" destructive (destructiveHint: true), and its name says "view", a word for reading.
What this check looks forMatched text
annotations: {"readOnlyHint":false,"destructiveHint":true,"openWorldHint":true}; The name "make_pose_view" says "view": a word for reading.
History
Changes to its tool definitions and to its remote endpoint, newest first.
Tools10 added, 2 changed.+10 added~2 changedView diff
- Added:add_wardrobe_itemChanges things
- Added:check_pose_designChanges things
- Added:check_pose_viewOnly reads
- Added:design_poseDeletes or cancels
- Added:finalize_video_draftDeletes or cancels
- Added:get_poseOnly reads
- Added:get_wardrobe_itemOnly reads
- Added:make_pose_viewDeletes or cancels
- Added:search_posesChanges things
- Added:search_wardrobeOnly reads
- Changed:generate_imageDeletes or cancelsdescriptioninputSchema
- Changed:generate_videoDeletes or cancelsdescriptioninputSchema
PostureRemote posture changed. Tools that change things, listed without credentials: 14 -> 20.View diff
- Tools that change things, listed without credentials: 14 to 20
Tools2 changed.~2 changedView diff
- Changed:generate_3d_modelDeletes or cancelsdescription
- Changed:generate_videoDeletes or cancelsinputSchema
Install
Pinned to version 1.0.0. Values in angle brackets are yours to fill in.
claude mcp add --transport http budgetpixel https://mcp.budgetpixel.com/mcpclaude mcp add --transport http budgetpixel https://mcp.budgetpixel.com/mcpRun in a terminal.
Claude Code permissions
Allow the reads, ask before a change or a way out, deny the deletes. Paste into .claude/settings.json.
{
"permissions": {
"allow": [
"mcp__budgetpixel__check_generation_status",
"mcp__budgetpixel__get_credit_balance",
"mcp__budgetpixel__list_models",
"mcp__budgetpixel__get_generation_history",
"mcp__budgetpixel__list_my_themes",
"mcp__budgetpixel__list_characters",
"mcp__budgetpixel__get_character",
"mcp__budgetpixel__search_wardrobe",
"mcp__budgetpixel__get_wardrobe_item",
"mcp__budgetpixel__get_pose",
"mcp__budgetpixel__check_pose_view"
],
"ask": [
"mcp__budgetpixel__save_to_assets",
"mcp__budgetpixel__upload_media",
"mcp__budgetpixel__recommend_influencers",
"mcp__budgetpixel__add_wardrobe_item",
"mcp__budgetpixel__search_poses",
"mcp__budgetpixel__check_pose_design"
],
"deny": [
"mcp__budgetpixel__generate_image",
"mcp__budgetpixel__generate_video",
"mcp__budgetpixel__finalize_video_draft",
"mcp__budgetpixel__generate_music",
"mcp__budgetpixel__generate_sound_effect",
"mcp__budgetpixel__generate_3d_model",
"mcp__budgetpixel__upscale_image",
"mcp__budgetpixel__upscale_video",
"mcp__budgetpixel__generate_depth_map",
"mcp__budgetpixel__search_stock_images",
"mcp__budgetpixel__search_stock_audio",
"mcp__budgetpixel__create_social_post",
"mcp__budgetpixel__design_pose",
"mcp__budgetpixel__make_pose_view"
]
}
}Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. A read that can send data out (it takes a URL, say) asks first too. It assumes the server is installed as budgetpixel. Review it before you use it.
Capability tags behind 17 rules
- mcp__budgetpixel__generate_image untrusted-input
- mcp__budgetpixel__generate_video untrusted-input
- mcp__budgetpixel__generate_music untrusted-input
- mcp__budgetpixel__generate_sound_effect untrusted-input
- mcp__budgetpixel__generate_3d_model untrusted-input
- mcp__budgetpixel__upscale_image untrusted-input
- mcp__budgetpixel__upscale_video untrusted-input
- mcp__budgetpixel__generate_depth_map untrusted-input
- mcp__budgetpixel__search_stock_images untrusted-input
- mcp__budgetpixel__create_social_post untrusted-input, outbound
- mcp__budgetpixel__list_my_themes untrusted-input
- mcp__budgetpixel__get_character private-data, untrusted-input
- mcp__budgetpixel__get_wardrobe_item untrusted-input
- mcp__budgetpixel__add_wardrobe_item untrusted-input
- mcp__budgetpixel__get_pose untrusted-input
- mcp__budgetpixel__design_pose untrusted-input
- mcp__budgetpixel__make_pose_view untrusted-input
Pin
Keep the decision: fail CI when the tools change, follow every change, or show a badge.
Pin in CI
Fail a build when the tools you reviewed are not the tools the server serves today.
curl -s "https://radar.anlyon.com/api/v1/servers/com.budgetpixel/mcp/verify?hash=7e474f49d5401363e2d68a108863f042c3ac3d1a2c56370995815a1319933638" | jq -e '.changed == false'Pinned hash 7e474f49d540, from Nov 10, 2026, 12:18 UTC.
Watch
Get told when its tools change: a new tool, a new description, a new class.
https://radar.anlyon.com/servers/com.budgetpixel/mcp/changes.xmlAn Atom feed. Works in any feed reader or Slack's RSS app.
Badge
[](https://radar.anlyon.com/servers/com.budgetpixel/mcp)Source: MCP Registry · repository. Synced Oct 21, 2026, 17:52 UTC.
budgetpixel.com MCP server: questions and answers
Answered from the scan above, not from the publisher's description.
Is the budgetpixel.com MCP server safe?
- No scan can promise that. What its definitions show: 31 tools, 20 of which can change, send or delete something, and 4 findings from fixed checks. Allow the reads, ask before each change, and deny the deletes; the permissions block on this page does that.
Can the budgetpixel.com MCP server delete data?
- Yes. 14 tools can delete, cancel or revoke something: generate_image, generate_video, finalize_video_draft, generate_music, generate_sound_effect, generate_3d_model and 8 more.
Can the budgetpixel.com MCP server do things that cannot be undone?
- No tool is classed as irreversible: none of them sends, pays, publishes or merges, going by their definitions.
What tools does the budgetpixel.com MCP server have?
- 31 tools: 14 destructive, 6 write, 11 read. generate_image, generate_video, finalize_video_draft, generate_music, generate_sound_effect, generate_3d_model, upscale_image, upscale_video and 23 more.
Does the budgetpixel.com MCP server need sign-in or OAuth?
- Its remote endpoint listed its tools to an anonymous client, without sign-in.
Does the budgetpixel.com MCP server send data to outside services?
- 1 tool can carry data out (it sends, posts or publishes, or takes a recipient, channel or URL): create_social_post.
Has the budgetpixel.com MCP server changed its tools?
- Yes, 2 times since Anlyon Radar started tracking it, most recently on Nov 10, 2026. Every change is on this page and in its Atom feed.