Agents & API
Your agent can
run your studio.
If you already talk to Claude while you work, it can do the busywork here too: create the gallery, add the finished photos, set the design, and tell you which clients have viewed, favorited, and sent their picks. Gallery Delivery connects to your agent through a token you create, with only the access you grant — nothing more.
Two ways in, one token: an MCP server at https://gallerydelivery.com/mcp
for agents like Claude, and a plain REST API under https://gallerydelivery.com/v1 for scripts.
How to connect
Three steps, about two minutes
-
Create a token
In your studio, open Settings → Security → API tokens, name the token (say, “Claude”), and tick the scopes it needs. The token itself — it starts
gd_live_— is shown exactly once, so copy it then. -
Paste the config
Give your agent the connection below with your token in place of
gd_live_YOUR_TOKEN. That's the whole setup — no passwords shared, no signing in as you. -
It works within its scopes
Your agent can only do what the token's scopes allow — that is enforced on every single request, not just suggested. Take a scope away or revoke the token and it stops, immediately.
Claude Code
Muse mcp add --transport http gallery-delivery https://gallerydelivery.com/mcp \
--header "Authorization: Bearer gd_live_YOUR_TOKEN"
Any MCP client — JSON config
{
"mcpServers": {
"gallery-delivery": {
"url": "https://gallerydelivery.com/mcp",
"headers": {
"Authorization": "Bearer gd_live_YOUR_TOKEN"
}
}
}
}
Through MCP your agent gets plainly-named tools: studio_info,
gallery_list, gallery_create, gallery_update_design,
gallery_set_access, photo_upload_from_url, photo_list,
and engagement_get. Ask it in plain words — “make a gallery for Saturday's
family session and add these photos” — and it uses them.
The scope model
You decide what your agent may touch
A token carries one or more scopes. Most studios start with read-only and add write access when they're ready — you can hold several tokens at once, each with its own mix.
| Scope | What it allows |
|---|---|
galleries:read | Read galleries — list and view galleries, photos, and studio info |
galleries:write | Create & edit galleries — details, design, and access settings |
photos:write | Upload & delete photos — including upload-from-URL |
engagement:read | Read engagement — favorites, picks, and comments |
REST API
Prefer scripts? It's plain HTTP
Every endpoint takes the same bearer token and answers JSON. Errors arrive as
{ "error": { "code", "message" } } with an x-request-id header
on the response, and each token may make 600 requests an hour. A few examples:
Create a gallery
curl -X POST https://gallerydelivery.com/v1/galleries \
-H "Authorization: Bearer gd_live_YOUR_TOKEN" \
-H "Content-Type: application/json" \
-d '{"title": "The Hanks Family — Newborn", "client_name": "The Hanks Family"}'
Add a photo from a URL
curl -X POST https://gallerydelivery.com/v1/galleries/GALLERY_ID/photos/from-url \
-H "Authorization: Bearer gd_live_YOUR_TOKEN" \
-H "Content-Type: application/json" \
-d '{"url": "https://example.com/finished-photo.jpg"}'
From-URL adds photos only — the fetched image itself becomes the display copy, and any proofing copies are prepared later from the gallery's manage page, the same as older photos. Videos need the upload endpoint, because a video also needs its poster frame.
Read a gallery's engagement
curl https://gallerydelivery.com/v1/galleries/GALLERY_ID/engagement \
-H "Authorization: Bearer gd_live_YOUR_TOKEN"
Safety
Built like a credential, because it is one
Shown once, hashed at rest
A token's secret is displayed a single time, at creation. Only its hash is stored — a database leak hands out no usable tokens, the same posture as your gallery access codes.
Scopes enforced per request
A read-only token that tries to change something is refused by the server, every time. The list you ticked is the whole of what the token can do.
Revoke instantly
One click in Settings → Security and the token dies on its very next request, everywhere. No waiting for anything to expire.
Every call logged
Each API and MCP call is recorded against the token that made it, so there's always an answer to “what did my agent just do?”
Honest limits
What isn't here yet
- No OAuth yet. Connecting means pasting a token you created — simple and fully in your control, but there's no one-click “connect” flow today.
- No video from a URL. Photos can be added from a link; videos go through the upload endpoint so each one arrives with its poster frame.
- 600 requests an hour per token. Plenty for an agent working a studio; a hard stop for a leaked token being scripted against us.
- Tokens are for your agent, not your clients. A token only ever reaches your own studio — it can't see another studio's galleries, and it never signs anyone into the studio pages.