Teach the library to recognize someone (an artist, presenter, team member) in every future upload. Add their name and 3–5 clear photos — different angles and settings work best.
Loading…
Recognition applies to images uploaded from now on. To find them in older images too, ask Claude to re-analyze (it costs a fraction of a penny per image).
Team API keys (for connecting Claude/MCP or scripts)
Each person or system gets their own key, so access can be revoked individually. Read-only keys can search and fetch URLs; full keys can also upload and edit.
Loading…
⚠️ Copy this now — it won't be shown again:
Master API key (system & local development only)
Team members sign in with their work email — you never need the key below.
Used by software that talks to the library (Claude/MCP, scripts) and when running the gallery outside the email login. Stored only in this browser.
Drop images here or
Each image is deduplicated, uploaded to Cloudflare, and auto-tagged by AI (~5–10s each).
How the Image Library works
The short version: drop images into Dropbox once → they're stored, optimized, and AI-tagged on Cloudflare → find them here by describing what you need → copy a URL or download. Dropbox is only the front door.
Do I need to keep the images in Dropbox? Can I make them "online only"?
Once an image has been ingested, Dropbox's job is done. The library keeps its own copy on Cloudflare — the gallery, search, URLs, and downloads never touch Dropbox again.
So yes: you can safely make the files Online Only (right-click in Finder/Explorer) to free up space on your computer. The gallery is unaffected. You do not need to download or sync anything to use this site.
⚠️ But think twice before deleting originals from Dropbox. The library serves optimized copies (resized, compressed for web/email) — not your untouched original files. If you delete the Dropbox original, the highest-quality master is gone. Online Only is the best of both worlds: no space used, original safely kept.
How do I add images?
Drop them into the shared Dropbox folder ISSPF Image Bank:
ISSPF Image Bank/Level 1 Coaching/photo.jpg → tagged with that course automatically
ISSPF Image Bank/Youth Academy/U12 Training/photo.jpg → tagged with brand and course
Within ~5 minutes each image is checked for duplicates, uploaded, and described + tagged by AI. The file then moves itself into _Processed/ — that's your receipt. Files that couldn't be ingested land in _Failed/ (usually iPhone HEIC format or over 10 MB — export as JPEG and re-drop).
Files you drop can be online-only — the system reads them from Dropbox's cloud, not from your computer. You can also drag-drop directly onto the Upload tab here.
How do I find images?
Type what you'd say to a person: "a player celebrating a goal", "a coach giving a team talk", "a packed stadium at night". The AI wrote a description of every image at upload, search finds the closest matches, then double-checks each result actually fits what you asked. Narrow further with the brand tabs and the course/person/type dropdowns.
New uploads appear in search a few minutes after ingesting — if something's missing, wait two minutes and search again.
How do I use an image in an email, on a website, or on socials?
Copy web URL — 1200px, for websites and social posts
Copy full URL — largest version, for hero banners
Download — saves the file to your computer
Paste the URL anywhere an image URL is accepted. The URLs don't end in .jpg — that's normal: they point at Cloudflare's image service, which automatically serves each visitor the lightest format their app supports. They're public links (protected only by being unguessable), which is exactly what email recipients and website visitors need.
Logging in — and that "API key" bar
You log in with your work email (@isspf.com) and a one-time code — no password, no key. If you ever see an "API key" bar, ignore it: it's for connecting software systems, never for people. Refresh the page and carry on.
Asking Claude for images (MCP)
The library is connected to Claude via MCP, so in a Claude session you can say things like "find me a photo from the coaching course and give me the email URL" or "how many images do we have per course?" — Claude searches this same library and pastes URLs straight into whatever it's writing (emails, landing pages, social posts).
Claude can also fix tags ("that second image is the wrong person — remove them") and check the Dropbox inbox. Connecting takes one minute and installs nothing: get a key from the gear icon → Settings → Team API keys, then in Claude add a custom connector with the URL https://images.isspf.com/api/mcp?key=YOUR_KEY (Claude Desktop: Settings → Connectors; Claude Code: claude mcp add --transport http).
To teach the library to recognize a new person (artist, presenter, team member): the gear icon → Settings → People recognition — name plus 3–5 photos, done in a minute.
Something's not working
Image didn't appear yet? Ingestion runs every 5 minutes, and search indexing takes ~2 more. Check _Failed/ in the Dropbox folder if it's still missing after that.
"Still being analyzed" in the counter? That's the system retrying automatically — no need to edit anything manually. It clears itself within the hour. Only "need attention" items want a human.
Wrong tags or person? Click the image → Edit, or ask Claude — every tag is editable.
Page looks broken? Hard refresh: Cmd/Ctrl+Shift+R.
Image Library