Connected AI tools
Cauldron speaks MCP (Model Context Protocol), the standard AI tools use to connect to outside services. Connect one and it can work inside your studio directly: search your Library, read your brand kits, compare models, and generate images that land in your Library like any other creation.
Nothing changes about how the studio works. A connected tool acts as you — same brands, same permissions, same credits — it just does it from wherever you’re already working.
What this is good for
Section titled “What this is good for”Once you’re connected, you talk to your AI tool in plain language and it reaches into Cauldron for you:
- “Find the hero shot from the summer launch project and upscale it.”
- “What brands do I have in Cauldron, and what’s in the Acme brand kit?”
- “Show me my last five generations.”
- “Which of our models is best for a poster with readable text on it?”
- “Polish this prompt, then make four 16:9 versions for the Acme brand.”
- “Cut the background out of the product shot I made yesterday.”
- “Import my Dropbox folder Product Shoots into the Acme brand.”
The results are real assets in your Library — they show up in the studio, in Review, and for your team, exactly like work you made in the composer.
Connect your AI tool
Section titled “Connect your AI tool”Every tool below connects by signing in with your Cauldron account, so there’s no key to copy. Wherever you start, it ends the same way:
- Your tool opens Cauldron in a browser. Sign in if you aren’t already.
- Pick the studio to connect, and click Allow.
The connection works as you, in the studio you picked. It’s listed on your profile under Connected AI Tools › Connected apps. Click Disconnect there and it stops working right away. Leaving the studio, or becoming a guest in it, cuts it off too.
Once you’re connected, ask something small to confirm, like “list my Cauldron brands”.
Claude Code
Section titled “Claude Code”The quickest way in is the Cauldron plugin. It connects the server for you and adds skills that teach Claude how to use it well: which model to pick, which brand to create into, and what costs money. The migrate skill walks Claude through importing a library from Dropbox. Run these two commands inside Claude Code:
/plugin marketplace add cauldronstudio/skills/plugin install cauldron@cauldronThen type /mcp, pick cauldron, and choose Authenticate. There’s nothing to fill in during the install.
Without the plugin, add the server from your terminal, then authenticate from /mcp the same way:
claude mcp add --transport http cauldron https://app.cauldron.studio/api/mcpAdd --scope user if you want Cauldron available in every project rather than just the current one.
Claude (claude.ai, Claude Desktop, and Cowork)
Section titled “Claude (claude.ai, Claude Desktop, and Cowork)”- In Claude, open Settings › Connectors and click Add custom connector.
- Name it Cauldron and set the URL to
https://app.cauldron.studio/api/mcp. Leave the advanced settings empty. - Click Add, then Connect, and sign in.
Connectors belong to your Claude account, so once it’s added on claude.ai it’s there in Claude Desktop and Cowork too. On a Claude Team or Enterprise plan, an owner adds the connector for the organization first; then each person clicks Connect and signs in as themselves.
Cursor
Section titled “Cursor”Add Cauldron to ~/.cursor/mcp.json (every project) or .cursor/mcp.json (one project):
{ "mcpServers": { "cauldron": { "url": "https://app.cauldron.studio/api/mcp" } }}Then open Cursor’s settings, find cauldron in the MCP list, and sign in when it asks. In the Cursor CLI, run cursor-agent mcp login cauldron instead.
codex mcp add cauldron --url https://app.cauldron.studio/api/mcpcodex mcp login cauldronThe second command opens the browser to sign in.
Skills for Cursor, Codex, and other agents
Section titled “Skills for Cursor, Codex, and other agents”The Claude Code plugin’s skills work in other agents too. npx skills add cauldronstudio/skills installs them. They don’t connect the server, so set up the connection above as well.
Other AI tools
Section titled “Other AI tools”Any MCP-capable tool works. Cauldron is a streamable HTTP MCP server at https://app.cauldron.studio/api/mcp. Add it with that URL and no header. If the tool supports MCP sign-in (OAuth), it opens the browser and you’re done.
If it doesn’t, use a key and send it as a header. Tools that take JSON configuration usually want a block shaped like this:
{ "mcpServers": { "cauldron": { "type": "http", "url": "https://app.cauldron.studio/api/mcp", "headers": { "Authorization": "Bearer oc_…" } } }}Check your tool’s own documentation for where that file lives and what it calls each field.
Use a key instead
Section titled “Use a key instead”Only tools that can’t sign in need a key, and you make it in the studio.
- Open your profile — your avatar menu, or Profile in the sidebar.
- Scroll to Connected AI Tools.
- Type a name that tells you which tool and which machine it’s for — My laptop, Zed — and click Create key.
- Copy the key now. It starts with
oc_and it’s shown exactly once. After you leave the page, only the first few characters remain visible.
Send it as an Authorization: Bearer oc_… header. Three things worth knowing about a key:
| It’s yours | Everything the connected tool does happens as you, with your permissions. Keys are personal — don’t share one with a teammate, give them their own. |
| It’s pinned to one studio | A key works only in the studio you were in when you created it. It cannot see another studio’s brands or assets, even one you belong to. Work across two studios? Make a key in each. |
| You can kill it anytime | Click Revoke next to a key and it stops working immediately. |
Working in more than one studio
Section titled “Working in more than one studio”A connection talks to exactly one studio: the one you picked when you signed in, or the one you were in when you made its key. There’s no studio picker and no switching inside a session, and nothing your AI tool sends can reach a different studio. Everything it makes lands in that one studio’s Library, on that studio’s credits, as you.
To work in a different studio, reconnect:
- Claude Code (plugin or not): type
/mcp, pick cauldron, choose Clear authentication, then Authenticate again and pick the other studio. - Other tools you signed in to: disconnect it on your profile under Connected apps, then sign in again from the tool and pick the other studio.
- Connected with a key: make a key in the other studio and replace the
Authorizationheader in your client’s config.
Keep it to one studio per AI tool. You can connect two studios as two separately named servers, but then both are live in the same conversation, and it’s your tool, not Cauldron, that decides which one a request goes to. Cauldron can’t blur them, but a misread request can land work in the wrong studio. One at a time, and the question never comes up.
What a connected tool can do
Section titled “What a connected tool can do”Seventy tools come across the connection. You never call them by name — you describe what you want and your AI tool picks — but it helps to know the shape of what’s available.
Find things
| Tool | What it does |
|---|---|
search | Finds Library assets, uploaded references, collections, templates, and imported files. Describe what’s in a picture in plain words (“ceramic candles in earthy colors”) and it finds it by meaning, not only by file name. Give it an asset and it finds ones that look like it, like Find similar in the Library. Put a date in the words (“hero shots from last week”, “latest renders”) and it searches that window, like search (⌘K). It covers what you can see in the Library: your own work and your teammates’ on brands you share. |
get_recent_generations | Your most recent creations, newest first — reaching back past today, not just the last 24 hours. |
show_generation | Returns the actual image so your AI tool can see it, not just read a description of it. |
get_template | Looks up a saved template by name or id — its steps, model, prompt, and settings. |
get_context | Orients the tool: who you are, which studio you’re in, and which brand you were last working in. |
Read your brands and models
| Tool | What it does |
|---|---|
list_brands | Every brand you can use here, with your role on each. Call it first: every write names a brand. |
get_brand | One brand’s full kit — prompt prefix and suffix, banned terms, palette, and anchors. |
list_models | Which models are available to you, what each is good for, and what each costs. |
get_model_details | The deeper read on one model: strengths, tradeoffs, and when to reach for it. |
Make and change images
| Tool | What it does |
|---|---|
generate_image | Generates 1–4 images and saves them to your Library. Takes a prompt and a brand, and optionally a model and aspect ratio. The brand’s kit applies exactly as it would in the composer, and the images start private to you, like any generation. |
enhance_prompt | Rewrites a rough prompt into a fuller, generation-ready one. It doesn’t generate anything on its own. |
upscale_image | Increases an image’s resolution and detail. |
remove_background | Cuts the subject out onto a transparent background. |
vectorize_image | Traces an image into vector art and saves it as an SVG that stays sharp at any size. Best on logos, icons and flat art. It costs 2 credits and turns down an image that’s already an SVG. Your AI tool sees a picture of the result; the link opens the SVG itself. |
extend_image | Grows an image past its edges and fills in what belongs there — a wider frame, room for a headline, a new aspect ratio without cropping the subject. |
resize_image | Resizes to exact pixel dimensions, with a choice of crop, pad, or stretch, and optionally converts the format. Nothing is redrawn, so it’s free and instant. |
All of them save a new image next to the original and never touch the source. If you want both on one image, upscale first — removing the background creates transparency, and upscaling would fill it back in with white. Ask for them the wrong way round and Cauldron says so rather than quietly ruining the cutout. Running the same one twice costs nothing extra either: you get the saved result back.
Make videos and GIFs
| Tool | What it does |
|---|---|
make_slideshow | Makes a short draft video (MP4, or a GIF up to a minute) from 2–60 images of one brand, from a collection or a list of images, in that order. You can set seconds per slide, crossfade, shape, fit, and MP4 or GIF. Add narration with a script and a narrator reads it over the slideshow (MP4 only); that spends credits. |
trim_video | Cuts a clip from a video in your Library and saves it as a new draft. Give it a start and an end in seconds; a clip runs from half a second to ten minutes. |
add_voiceover | A narrator reads a script over a video in your Library, saved as a new video; the original stays. Choose warm, clear, bright, friendly, firm or gentle, or leave it out and the brand’s narrator reads it. Spends credits for the script’s length; your AI tool should show you the script, voice and price first. |
make_web_ready | Turns a heavy GIF or video into a small, muted MP4 and WebM with a poster, saved as a new version of the same asset. |
These are the same tools as in the studio, with the same rules: you need to be able to create in the brand, and studio Guests can’t run them. Each runs in the background, so your AI tool gets a link to Jobs to follow it, and the result lands in your Library.
Run a template
| Tool | What it does |
|---|---|
run_template | Runs a saved template the way the Templates page does: name the template, give it the prompt or images its steps ask for, and name the brand its results land in. It spends credits, so your AI tool asks before it starts. It hands back a run id, the run goes on in Jobs, and every result lands in your Library. |
get_template_run | Checks on a run: which step it’s on, whether it’s waiting for you to check an image, finished, or failed, and what each step made. Only the person who started the run, or a studio owner or admin, can read it. |
continue_template_run | Resumes a run that paused, like Animated hero waiting for you to check the image, or one that failed, so it goes on to the next step. Only the person who started the run can resume it, the same rule as the run page. You can also ask it to skip the run’s later pauses. |
Add your own files
| Tool | What it does |
|---|---|
upload_asset | Puts a file into a brand’s Library, from a web link or straight from your AI tool’s own disk (up to 10 MB that way), and tags it as it lands. Same file types and limits as uploading in the studio. |
presign_uploads | Hands out upload links for up to 1,000 files on your AI tool’s disk at once, for a migration or a big batch. Your AI tool sends each file’s bytes straight to storage. Same file types and limits as uploading in the studio; files it can’t take are listed with the reason. Files over 64 MB are sent in parts, so a dropped connection only resends one part. |
complete_multipart_upload | Finishes a large file your AI tool sent in parts. |
abort_multipart_upload | Cancels a large file your AI tool sent in parts and deletes what was uploaded. |
import_assets | Files what presign_uploads sent into a brand’s Library, up to 1,000 at a time, under one collection with nested collections inside it, plus tags, description, rating, details and comments for each file. It runs in the background: follow it with get_import_status. A file that was imported into the brand before is linked again rather than copied, so running a migration twice adds nothing. |
This is what makes the image tools useful on files you already have: upload a logo, then resize it to an exact size, upscale it, or use it as a reference for a generation. Only people who can create in the brand can upload to it, and adding the same link twice gives you back the asset you already have rather than a duplicate.
Edit what’s already in your Library
| Tool | What it does |
|---|---|
suggest_asset_names | Suggests a few names for an asset from what’s in it and how the brand names things, so you can pick one before update_asset renames it. |
update_asset | Changes an asset’s tags, title, or description. Add and remove tags, or replace the whole set. Tag names match the tags you already have, ignoring case and punctuation, so “Spring Sale” and “spring-sale” land on the same tag; a name that’s new becomes a tag for the whole studio, the same as typing it in the Library. |
You can edit an asset here when you could edit it in the Library: creators and brand managers on the asset’s brand, and studio owners and admins. Ask for the asset by name or describe it, and your AI tool finds it with search first.
Tidy up your tags
| Tool | What it does |
|---|---|
list_tags | Lists the tags you can see, as on the Library’s Tags page: each tag’s brand, category, how many assets carry it, and whether it’s an AI suggestion nobody has reviewed yet. Narrow it to one brand or a name. |
delete_tag | Deletes a tag and removes it from every asset that carries it. This can’t be undone, so your AI tool should check how many assets use the tag with you first. |
You can delete a tag when you could delete it on the Tags page: brand managers for their brand’s tags, and studio owners and admins for every tag.
File into collections and projects
| Tool | What it does |
|---|---|
list_collections | Lists the collections you can see, as the Library rail shows them, narrowed to one brand when you name it. |
create_collection | Makes a new collection on a brand. Studio-wide collections stay a job for the Library. |
update_collection | Renames a collection or changes its description. The same name and description again changes nothing. |
add_to_collection | Files assets that are already in your Library into a collection. Assets that are already there are left as they are, and the tool says so. |
remove_from_collection | Takes assets out of a collection. They stay in your Library. |
delete_collection | Deletes a collection after your AI tool asks. The assets in it stay in your Library. |
list_projects | Lists projects with their start and due dates and how many assets are approved, in review, or still drafts. Active ones by default; ask for archived or all, one brand’s, or studio projects. |
get_project | Opens one project by name or id: its dates, approvers, counts, its newest assets and, if you manage it, its share link. A name that matches more than one project gets you the candidates to pick from. Archived projects open too. |
create_project | Starts a project on a brand, with a start and due date if you give them. Studio owners and admins can start a studio project instead. |
update_project | Moves a project’s dates or clears them, archives it, or brings it back. An archived project takes no other change until it’s back, a due date can’t be before the start, and renaming stays on the project page. |
You can change a collection here when you could in the Library: studio owners and admins, and creators on the collection’s brand. Guests can’t. Filing an asset also counts as editing it, so you need to be able to edit the asset too. Projects follow the Projects page: guests see only their brands’ projects and never studio projects; a brand Creator or Manager, or a studio owner or admin, changes a brand’s project, and only an owner or admin a studio project.
Saved views
| Tool | What it does |
|---|---|
list_views | Lists the saved views you can use: yours and the ones shared with the brand, each with its filters in plain words and a Library link. |
apply_view | Runs a view: the count the Library shows, its newest matches, and a link to open it in the Library. |
save_view | Saves a search, filters and dates as a new view, private to you, on the brand you name. It can save what you have open in the Library. |
update_view | Renames a view, replaces its filters, or shows it as a smart collection on the Collections page. |
delete_view | Deletes a view after your AI tool asks. It never touches your files. |
Products
| Tool | What it does |
|---|---|
list_products | A brand’s products: each one’s name, what it is, confirmed photos, whether it’s trained, open issues, and the images pushed to its store page. |
get_product | One product in full, including the approved images that aren’t on its store page yet, so your AI tool can offer to push them. |
Share links
| Tool | What it does |
|---|---|
list_shares | Lists a brand’s share links, newest first, as Settings → Shared links shows them: what each one shows (a collection, project or asset), its link, whether it shows approved work only, whether downloads are on, whether it has a password, its note, who made it, how many times it was opened, and when it expires. Expired links stay on the list. Ask for only the links you made with mine, or for one collection’s or asset’s link. |
get_share | Reads one link: what it shows, its settings, who made it and how many times it was opened. |
create_share | Publishes a collection, project or single asset as a view-only link and returns its URL. A new link expires after 30 days unless you ask for another date or no expiry, and it can have a password. If the brand allows mature content, your AI tool has to ask whether you want a password before it can publish. |
update_share | Changes a link’s expiry, password, downloads, approved-only setting or note. The URL stays the same. |
revoke_share | Turns a link off for good. Everyone who has the URL loses access, and sharing again makes a new URL. |
list_shared_collection | Lists every asset on a collection’s share link with a preview link (a small image that works in Google Sheets’ =IMAGE()) and, once the asset is approved, a download link. The collection must already be shared, and you must be able to manage its share link. It never creates or changes a link. |
You need a creator or manager role on the brand for every share tool, and a connection with write access to create, change or turn off a link. Creating a link publishes your work to anyone who has the URL, so your AI tool should only do it when you ask, and should give the URL to you rather than post it anywhere.
The preview and download links keep working until the share link is turned off or reset, and anyone who has them can open them.
Send to a store or a social scheduler
| Tool | What it does |
|---|---|
list_destinations | Lists the stores and social schedulers connected to your brands, what each one takes (images, video), whether you can send to it and whether you can schedule through it. Give it a few words (“the Instagram profile”, “amber candle”) and it also finds profiles and products inside them. If nothing is connected, it says so. |
send_to_destination | Sends approved assets to a store’s product or files, or to a social profile. A social send is saved as a draft post in the scheduler with the caption you give it. With a time, it schedules the post instead, if a brand manager has turned on Agents can schedule. Each asset that can’t go comes back with the reason. |
list_deliveries | Says how each send ended: still sending, sent, failed and why, or removed. Pass an asset to see every send of it, the sends send_to_destination just made, or every send through one destination. Narrow by profile, by start time, or by a post’s state: draft, scheduled, published or didn’t publish. For a social post, the result carries the scheduler’s own post id. |
cancel_delivery | Cancels one draft or scheduled social post: it’s deleted in the scheduler, and the asset can be sent there again. A published post can’t be canceled. |
Only approved assets go. A connected tool can’t approve a draft on the way out, can’t send an asset marked restricted (a person can send it from Cauldron), and sends at most 10 assets to a social profile at a time. You need a creator or manager role on the brand and a connection with write access to send. Sending puts your work outside Cauldron, so your AI tool should only do it when you ask.
A scheduled post goes live without anyone opening Cauldron, so scheduling is refused until a brand manager turns on Agents can schedule on the Vista Social row; your tool should offer a draft instead. The time needs a UTC offset, at least 15 minutes ahead and within a year. If the brand is narrowed to a profile group, only that group’s profiles can be found or sent to.
Connecting a store or Vista Social happens in Cauldron, on the brand’s Integrations tab. A brand manager connects it; a connected tool can’t.
Pull stills out of a video
| Tool | What it does |
|---|---|
extract_frames | Saves full-size stills from one video or GIF as drafts in a new “Frames from …” collection in the same brand. Ask for a number of frames spread across the clip, one every few seconds, or exact moments. Up to 60 frames a run, from videos up to 10 minutes or 500 MB. It runs in the background and shows in Jobs. |
You need a creator or manager role on the brand and a connection with write access. It’s free: no credits.
Bring a library in from Dropbox
| Tool | What it does |
|---|---|
dropbox_status | Whether your Dropbox is connected, which account, and any import that’s running. |
dropbox_browse | Lists one folder level of your Dropbox, so an agent can pick folders with you before anything is counted. |
dropbox_scan | Counts only the folders you picked, so they can be previewed. |
dropbox_get_scan | Lists folders with their file counts, one level at a time. |
dropbox_preview | Shows what an import of the folders you chose would do, before anything is copied. |
dropbox_import | Starts the import. |
get_import_status | How far an import has got (from Dropbox or import_assets), which files failed and why, and where the files landed. |
resume_import | Retries the files that failed or are still waiting. |
cancel_import | Stops an import. Files that already arrived stay. |
These cost nothing. Connecting Dropbox itself happens in Cauldron, not in your AI tool: see Import with a connected AI tool.
Bring a library in from pics.io
| Tool | What it does |
|---|---|
picsio_status | Whether you’ve connected pics.io on the brand. If not, it gives you a link to connect in Cauldron. Without the pics.io API, use presign_uploads and import_assets instead. |
picsio_scan | Reads the brand’s pics.io library, its collections and files, in the background. |
picsio_get_scan | Lists collections with their file counts, one level at a time. |
picsio_preview | Shows what an import of the collections you chose would do, before anything is copied. |
picsio_import | Starts the import. Follow it with get_import_status, and use resume_import or cancel_import as for Dropbox. |
These cost nothing. Connecting pics.io happens in Cauldron, with a pics.io access token, not in your AI tool: see Import from pics.io.
Studio and brand settings
| Tool | What it does |
|---|---|
get_settings | Shows the settings you can change: the studio’s (owners and admins only), each brand you manage, and whether Slack is connected. |
update_brand_settings | Changes one brand’s settings: name, description, website, location, self-approval, its Slack review channel, video, community publishing, prompt prefix and suffix, banned terms, and the narrator its voiceovers use. |
update_studio_settings | Changes the studio name or the Slack channel for studio updates. |
The same people who can change these in the studio can change them here: studio owners and admins, and brand managers for their own brands. Everyone else is told no.
Saving a Slack channel sends a test message to it, just like saving it in the studio, and nothing is saved if that message can’t be delivered. Invite the Cauldron bot to the channel first (/invite @Cauldron) and use the channel ID from the channel’s About tab, not its #name. Connecting Slack itself happens in Cauldron, in Studio settings › Slack; a connected tool can’t connect or disconnect it.
What spends money
Section titled “What spends money”A few tools do real work that costs something: generate_image, enhance_prompt, upscale_image, remove_background, vectorize_image, extend_image, and run_template. They only run when your studio’s billing is active, and an image made with generate_image or a template run draws down your studio’s credits exactly as it would in the studio. Everything else is free.
Every call, reads included, counts toward your plan’s monthly MCP allowance (200 on Starter, 1,000 on Solo, 10,000 on Pro, unlimited on Agency; see Billing). Calls are never charged one by one. Once the allowance is used up, your AI tool is told which plan lifts the limit, and calls work again when the plan renews.
Your AI tool’s own permission prompt is the confirmation step. When Claude Code (or Cursor, or anything else) asks “allow generate_image?”, that’s your one chance to say no — Cauldron doesn’t ask a second time. If you’d rather be asked every time, don’t set these tools to always-allow in your client.
Generations made this way spend credits just like generations made in the studio, and they show up in your usage and cost totals.
What isn’t available over MCP
Section titled “What isn’t available over MCP”Two things stay in the studio on purpose:
- Video generation. Video runs as a long job you start and come back to. That doesn’t fit a connected tool sitting and waiting for an answer, so make video in the studio — see Creating video.
- Reframe. Reframe works across a batch and tells you when it’s finished. “I’ll let you know” means nothing outside the app, so run it from the tools instead.
Your permissions come with you
Section titled “Your permissions come with you”A connected tool is not a way around anything:
- It only sees the studio you connected it to.
- It respects your role on each brand. View-only on a brand means it can read that brand’s kit but can’t create into it.
- Every generation, upload, and import goes into a named brand. If you can add to exactly one brand, that brand is used. Otherwise the tool is asked to pick one and given the list from
list_brands. - In a studio with no brand yet, writes are refused with a message to open the app and create your first brand.
- It spends your studio’s credits and counts against any monthly allowance you have on the brand.
- Revoking the key, or leaving that studio, cuts it off immediately.
Revoke a key
Section titled “Revoke a key”Open your profile, find the key under Connected AI Tools, and click Revoke. It stops working right away — no waiting, no restart. The row shows each key’s prefix, when you created it, and when it was last used, so you can tell which one is which before you revoke.
Revoke a key if a laptop goes missing, if you pasted the key somewhere it shouldn’t be, or if you’re just done with that tool. Then create a fresh one and reconnect.
Troubleshooting
Section titled “Troubleshooting”| What you see | What’s going on |
|---|---|
| Every tool fails with 401 or unauthorized | The key is wrong, revoked, or you’re no longer a member of that studio. Create a new key and update the header. Connected by signing in? Reconnect from your tool and sign in again (in Claude Code: /mcp → cauldron → Authenticate). |
| The plugin asks for an API key or a studio URL | You have an old version of the plugin. Run /plugin marketplace update cauldron, then /plugin update cauldron@cauldron, and sign in with /mcp. |
| Sign-in shows You can’t connect AI tools yet | You’re a guest in every studio you belong to. Guests can’t connect AI tools; ask a studio admin to upgrade your access. |
401 and your server URL still says studio.opencauldron.ai | The studio moved to app.cauldron.studio. The old address redirects, but your AI tool drops the key when it follows a redirect to a different host. Change the URL to https://app.cauldron.studio/api/mcp; the key itself still works. Plugin users: run /plugin marketplace update cauldron, then /plugin update cauldron@cauldron. |
| billing_inactive on generate, enhance, upscale, or remove background | Your studio’s billing isn’t active. Reads keep working; the four spending tools don’t. |
| A tool you know shipped isn’t there | Clients cache the tool list from when they connected. Restart or reconnect your AI tool to pick up new ones. |
| that model is async-only or video models are not supported | generate_image returns the finished image right away, so it only works with models that generate immediately. Ask your tool to list models and pick another. |
| An image link stopped loading | Image links are short-lived on purpose. Ask to see it again and a fresh link is issued. |
| The tool says it needs a brand | You can add to more than one brand and didn’t name one. Tell it which, or ask it to list your brands. |
| This studio has no brand yet | Open the app and create your first brand, then try again. |
| You have view-only access to this brand | Your role on that brand doesn’t allow creating. See Brands. |
| That image no longer exists | Either it’s gone, or it lives in a different studio than the key. Keys never cross studios. |
Related
Section titled “Related”- Your profile — where keys are created and revoked
- The Library — where everything a connected tool makes lands
- Brands — brand kits, roles, and permissions
- Usage & limits — credits and monthly allowances
- Editing images — Upscale, Remove background, and the rest, in the studio