# Tagrly full reference

> Tagrly turns a business's own photo library into an API an AI agent can call. A customer connects a Google Drive or Dropbox folder; Claude vision reads every image into 34 structured fields; an agent then POSTs a plain-English page topic to `/api/page` and gets back a page-ready set of that business's own photographs with alt text already rewritten for the topic, plus an honest gap signal naming the shots the library does not contain. One endpoint, `POST /api/try/analyze-photo`, takes no account at all. Tagrly is for putting a customer's own photographs into generated pages, posts, listings and decks. It is not a stock-photo service and it does not generate images.

This is the complete reference in one file. The short index is at https://tagrly.com/llms.txt; the same content as a web page is at https://tagrly.com/agent-guide. Last reviewed against the running code: 2026-08-04.

Base URL: `https://tagrly.com`

---

## 1. What you can do without an account

`POST /api/try/analyze-photo` is the only endpoint that takes no authentication.

Send one image as `multipart/form-data` under the field name `photo`. It runs the
same prompt and JSON schema a paying workspace gets, returns the structured
reading, and persists nothing: no session, no disk write, no database row. The
bytes live in the request and are dropped with it.

```
curl -X POST https://tagrly.com/api/try/analyze-photo \
  -F "photo=@kitchen.jpg"
```

200 response:

| field | meaning |
| --- | --- |
| `ok` | Always `true` on a 200. |
| `vertical` | Which vertical the photo classified into. Decides the schema that ran. |
| `prompt_version` | Version tag for the prompt that produced this reading. |
| `analysis` | The structured reading. Field list depends on `vertical`. |
| `elapsed_ms` | Server-side time for the classify and analysis calls. |
| `remaining` | Reads left in the current hour. |
| `limit` | Reads allowed per hour, so you can pace without guessing. |

Limits: 5 reads per hour per connection on a sliding one-hour window; 8MB per
upload, refused on the `Content-Length` header before the body is read; formats
JPEG, PNG, WebP, HEIC, HEIF, AVIF, GIF. Anything else that happens to decode is
rejected by name.

Errors: every body carries an `error` slug to branch on and a human-readable
`message` you can show a user verbatim:

| status | error | cause and response |
| --- | --- | --- |
| 400 | `no-file` | No part named `photo`. Fix the field name; do not retry as-is. |
| 400 | `empty-file` | Zero bytes arrived. Re-read the source file. |
| 400 | `wrong-type` | Decoded, but outside the format allowlist. The message names the format found. |
| 400 | `bad-image` | Would not decode at all. Different file. |
| 413 | `too-large` | Over 8MB. Re-export smaller and resend. |
| 429 | `rate-limit` | Hourly budget spent. `retry_after` holds the seconds to wait. |
| 500 | `server-misconfigured` | Server-side. Retrying will not help. |
| 502 | `parse-failed`, `upstream` | The model call failed or returned an unusable shape. Retry once, then back off. |

---

## 2. What one photo returns

The schema is composed at request time: a universal base plus an overlay for the
detected vertical, so the exact field list depends on the returned `vertical`.
The base fields are present on every reading, and they are also the fields the
catalog indexes, which is why search can filter on scene, mood or focal subject
later.

| group | fields |
| --- | --- |
| Writing | `alt_text`, `description`, `visual_summary`, `suggested_filename` |
| Subject | `focal_subject`, `focal_category`, `subject_tags`, `activities`, `branded_items`, `keywords` |
| Setting | `scene`, `mood`, `lighting`, `time_of_day`, `season_cues`, `visual_temperature`, `dominant_colors` |
| Frame | `shot_type`, `orientation`, `people_count`, `engagement_with_camera`, `has_negative_space`, `crop_friendliness`, `background_cleanliness` |
| Fitness for use | `quality`, `usability_rating`, `editorial_fit`, `is_marketing_graphic`, `text_visible`, `text_overlay_present`, `alcohol_prominent` |
| Safety | `minors_visible`, `privacy_concern_flags`, `faces_clearly_identifiable_count` |

Seven verticals classify: `universal`, `hospitality-sports`, `wedding-venue`,
`real-estate-listing`, `restaurant-bar`, `ecommerce-product`,
`automotive-listing`. Each adds its own fields, a wedding photo gets ceremony
phase, a sports photo gets jerseys and what is on the TV.

---

## 3. Authentication

Every endpoint except `POST /api/try/analyze-photo` takes one header:

```
Authorization: Bearer tagrly_pk_<workspace-slug>_<32 hex chars>
```

The random half is 128 bits. The workspace slug is visible on purpose, so a
customer's own logs show which tenant a call belonged to.

Key issuance is not self-serve. There is no signup flow, no settings screen and
no API endpoint that mints a key. A person at Tagrly generates it and sends it
to the customer, who emails support@tagrly.com from their account address. If
you are an agent and your user has no key, say so and ask them for one. Do not
guess at a key format or an issuance route.

Only the SHA-256 hash is stored, so the visible key is shown once and cannot be
recovered; a lost key is replaced, not read back. Revoking a key stops it
resolving on the very next request.

Every key grants full read/write access to its workspace. There is no
scope system yet, so do not branch on a scope field until this file says
otherwise.

One shape worth knowing: an anonymous request to a keyed endpoint does not
always return 401. Signed-out browser traffic is redirected to sign-in before
the endpoint runs, so a client that follows redirects can receive an HTML page
with a 200. Send the `Authorization` header on every keyed call and the question
never arises.

---

## 4. The core loop

1. Decide which section of the page you are filling, hero, gallery, room, detail.
2. `POST /api/page` with a plain-English topic for that section.
3. Render each pick's `hosted_url` with the `alt_rewrite` Tagrly wrote for it.
4. Post the rendered `drive_id`s to `/api/usage`.
5. Pass `exclude_recently_used_days` on the next call, so a long automation never repeats an image.

If you find yourself writing slot definitions by hand, you want `/api/brief`.
Most calls should be `/api/page`: it infers the layout and returns the same
response shape.

---

## 5. Every endpoint

| if you want to… | call | key |
| --- | --- | --- |
| Read one photo you upload | `POST /api/try/analyze-photo` | no |
| Fill a page section from a plain topic | `POST /api/page` | yes |
| Define slot shape yourself | `POST /api/brief` | yes |
| Log the images you rendered | `POST /api/usage` | yes |
| List what was logged lately | `GET /api/usage/recent` | yes |
| Run a keyword and facet search | `GET /api/search` | yes |
| Get one image's full analysis | `GET /api/image/{drive_id}` | yes |
| List named collections | `GET /api/collections` | yes |
| Save a set for human review | `POST /api/add-to-collection` (form-encoded) | yes |
| Resolve ids to source URLs | `GET /api/urls` | yes |
| Get every id matching a search | `GET /api/match-ids` | yes |
| List workspaces this key reaches | `GET /api/workspaces` | yes |

Bodies are JSON except two: `POST /api/try/analyze-photo` is
`multipart/form-data` (the image goes in a field named `photo`), and
`POST /api/add-to-collection` is `application/x-www-form-urlencoded`.

---

## 6. POST /api/page

The smallest useful request is one field.

```
POST /api/page
Authorization: Bearer tagrly_pk_acme_...
Content-Type: application/json

{ "topic": "weekend brunch on our rooftop patio" }
```

### Request fields

Only `topic` is required.

| field | default | what it does |
| --- | --- | --- |
| `topic` | n/a | Plain-English description of the section. An empty one returns `topic_required`. |
| `h1` | `""` | The page's headline. Anchors the alt rewrites and nudges toward images with negative space. |
| `vocabulary` | `{}` | Object with `extra_terms` and `teams` arrays. Use for words the catalog would not know. |
| `exclude_image_ids` | `[]` | Ids already used on this page by earlier calls. |
| `exclude_recently_used_days` | `0` | Skips anything logged via `/api/usage` in the last N days for this workspace. |
| `include_promo_graphics` | `false` | Set true when branded promo art is the point, e.g. a campaign landing page. |
| `safety` | `"safe"` | Safety filter mode applied to the candidate pool. |
| `dry_run` | `false` | Skips model curation. Inspect the candidate pool at zero cost. |
| `model` | `claude-haiku-4-5-20251001` | Curation model. A larger model buys sharper judgement at higher cost. |
| `cost_cap_usd` | `0.50` | Circuit breaker for one call. Clamped to a hard ceiling of `5.00`. |

### Response

```
{
  "ok": true,
  "topic": "weekend brunch on our rooftop patio",
  "layout": "food-drink",
  "slots": {
    "hero":         { "wanted": 1, "got": 1, "picks": [ ... ] },
    "product_grid": { "wanted": 6, "got": 6, "picks": [ ... ] },
    "experience":   { "wanted": 3, "got": 3, "picks": [ ... ] }
  },
  "gap_signal":  { "fired": false, ... },
  "diagnostics": { "claude_cost_usd": 0.029, "elapsed_ms": 9021, ... }
}
```

`layout` and `slots_inferred` are the two fields `/api/page` adds over
`/api/brief`; everything else is identical between them. Each slot also reports
`candidates_considered`, `shortlisted` and an `on_topic_score_distribution`,
which together tell you whether a thin result means a thin library or a narrow
topic.

Top-level fields also include `workspace_id`, `workspace_name`,
`expanded_terms` and `auto_discovered_aliases`.

### Fields on every pick

| field | use |
| --- | --- |
| `drive_id` | Stable id. Send this to `/api/usage` and `exclude_image_ids`. |
| `hosted_url` | Public URL. Render directly in `src`. |
| `thumb_url` | Smaller variant for previews and pickers. |
| `alt_rewrite` | Alt text rewritten for this page's topic. `null` when `alt_text_rewrite` is off. |
| `alt_text` | The catalog's own alt text, independent of this request. |
| `on_topic_score` | 0–100. Under 50 is tangential; decide deliberately whether to use it. |
| `quality_score` | 0–100, from the analyzer's quality reading. |
| `prescore`, `prescore_signals` | Pre-curation ranking score and the signals that produced it. |
| `why_picked` | One sentence on why the curator chose it. |
| `focal_subject`, `focal_category` | What the image is of. |
| `scene`, `mood`, `shot_type` | Setting and framing, for your own grouping. |

### Diagnostics

`diagnostics` carries `candidates_pool_total`, `claude_calls`,
`claude_input_tokens`, `claude_output_tokens`, `claude_cost_usd`,
`cost_cap_usd`, `cost_cap_hit`, `recently_used_excluded`,
`exclude_recently_used_days`, `time_context_detected`,
`promo_graphics_filtered`, `elapsed_ms`, `model_used`, `dry_run` and
`validation_warnings`.

### Layout detection

First match wins, and matching happens within the workspace's vertical, a food
topic only reaches the food layout in a hospitality workspace.

| when the topic… | layout | slots |
| --- | --- | --- |
| Names an entity the workspace defined, or passes `vocabulary.teams` | `sports` | hero, action_grid, lifestyle |
| Carries food or drink words, in a hospitality vertical | `food-drink` | hero, product_grid, experience |
| Carries celebration or private-event words | `events` | hero, moments_grid, details |
| Carries venue, room or space words | `venue` | hero, tour_grid, people_in_space |
| Matches nothing above | `generic` | hero, body_grid, detail |

---

## 7. POST /api/brief

Same pipeline, your slots. Reach for it when the auto layout does not match your
template. Every `/api/page` field applies, plus three:

| field | default | what it does |
| --- | --- | --- |
| `slots` | n/a | Array of slot objects. An empty array returns `slots_required`. |
| `alt_text_rewrite` | `true` | Off returns `alt_rewrite: null` and skips that part of curation. |
| `max_candidates_per_slot` | `25` | How deep the shortlist goes before curation. Higher costs more. |

```
POST /api/brief
{
  "topic": "our outdoor patio brunch",
  "h1":    "Weekend Brunch on the Patio",
  "slots": [
    { "role": "hero", "count": 1,
      "must_have": ["focal:food-or-drink"],
      "prefer":    ["quality>=75", "negative_space"],
      "diversity": "none",
      "description": "Lead photo, drink and dish together if possible." },
    { "role": "gallery", "count": 8,
      "must_have": ["focal:food-or-drink|focal:product-or-object"],
      "prefer":    ["quality>=70"],
      "diversity": "subject+angle",
      "description": "Eight distinct dishes. Vary the angles." },
    { "role": "atmosphere", "count": 4,
      "must_have": ["people>=2"],
      "prefer":    ["mood:lively|mood:celebratory"],
      "diversity": "subject+angle",
      "description": "Group moments on the patio." }
  ]
}
```

The response shape is identical to `/api/page`, same picks, same gap signal,
same diagnostics.

---

## 8. The slot language

Two lists per slot. `must_have` is a hard filter and every item must pass.
`prefer` is soft scoring: matches earn a bonus, misses are not excluded. Within
one string, alternatives join with `|`.

| predicate | true when |
| --- | --- |
| `focal:<a>,<b>` | The image's focal category is one of the listed values. |
| `scene:<a>,<b>` | Its scene is in the set, e.g. `scene:bar-area,outdoor-patio`. |
| `mood:<a>,<b>` | Its mood is in the set. |
| `people>=N`, `people>N`, `people<N` | People in frame compares as stated. |
| `quality>=N` | Quality score is at or above N. |
| `negative_space` | The image has clear space for a text overlay. |
| `no_back_of_head` | Neither alt text nor description describes a back-of-head shot. |
| `pillar:N` | The image is tagged into content pillar N. |
| `jerseys` | Sports overlay. Team jerseys were detected. |
| `team:<name>` | The team appears anywhere, jerseys, branded items, a broadcast on screen. Most permissive. |
| `team_focal:<name>` | The team is in the focal subject, alt text or jerseys. Excludes background-only mentions. |
| `team_dominant:<name>` | Strictest: the team leads the jersey list or the focal subject. |
| `sport:<name>` | The sport is on screen, or appears in the broadcast content or keywords. |

Diversity strategies. An unrecognised value falls back to `subject+angle`.

| strategy | effect |
| --- | --- |
| `none` | No rerank. Best-scoring images, in order. Use for single-image slots. |
| `subject+angle` | Round-robin across scene, shot type and focal category. The default, and right for most grids. |
| `scene` | Bucket by scene only. Maximises location variety. |
| `shoot` | Bucket by source folder. Avoids a grid that is visibly one shoot. |

---

## 9. The usage ledger

This is what stops a fifty-post automation from using the same hero fifty times.

```
POST /api/usage
{
  "items": [
    { "drive_id": "1xabcd...",
      "page_url": "https://customer.com/brunch",
      "page_topic": "weekend brunch",
      "slot_role": "hero" }
  ]
}
```

A single use may also be posted as a bare object without `items`. Only
`drive_id` is required.

| call | parameters | returns |
| --- | --- | --- |
| `POST /api/usage` | `items` array, or a single object. Batches over 100 are truncated to 100; entries with a malformed `drive_id` are skipped. | `{ ok, recorded, skipped }` |
| `GET /api/usage/recent` | `days`, default 14, max 365. `limit`, default 100, max 500. | `{ ok, count, days, items, drive_ids }` |

`drive_ids` comes back de-duplicated and sorted, which is exactly the shape
`exclude_image_ids` expects.

`/api/usage` inserts a row per item every time and does not de-duplicate. The
same image legitimately used twice on one page writes two rows, and the
freshness filter handles that.

---

## 10. GET /api/search

The primitive under everything else: keyword and facet search with no model in
the loop, so it is free and fast. Returns `{ count, results }`. Each result
carries the catalog fields, but not `alt_rewrite`, `on_topic_score` or
`why_picked`, because nothing curated it.

| parameter | accepts |
| --- | --- |
| `q` | Free-text query across the catalog's indexed fields. |
| `scene`, `mood`, `shot_type`, `focal_category`, `editorial_fit` | Exact facet values. |
| `folder` | Restrict to one source folder. |
| `source` | `drive` or `dropbox`. |
| `neg_space` | `yes` or `no`. |
| `pillar` | Integer 1–5. |
| `team`, `year` | Team name; four-digit year. |
| `safety` | `safe` (default) or `any`. |
| `dupes` | `hide` (default) or `show`. |
| `focal` | `1` to match against focal subject only. |
| `limit` | Default 24. Clamped to 500. |
| `sort` | Result ordering, e.g. by relevance or date. |

---

## 11. The remaining endpoints

| call | input | returns |
| --- | --- | --- |
| `GET /api/image/{drive_id}` | Path id, 5–100 characters of `A–Z a–z 0–9 _ - :` | The full analysis payload for one image. 404 when the id is not in this workspace. |
| `GET /api/collections` | none | Summary of this workspace's named collections. |
| `POST /api/add-to-collection` | Form-encoded `name` and `drive_ids` (comma-separated string) | `{ collection_id, collection_name, added }`. Matches an existing collection by name, case-insensitively, or creates one. |
| `GET /api/urls` | `ids`, comma-separated. Over 500 is truncated. | `{ urls }`, source URLs for ids present in this workspace. |
| `GET /api/match-ids` | The same filters `/api/search` takes | `{ ids, total, capped, cap }`. `capped` tells you the match set was larger than the 500 returned. |
| `GET /api/workspaces` | none | Workspaces this credential reaches, and which one is active. |

---

## 12. Defaults applied on every curated call

| filter | behavior | opt out |
| --- | --- | --- |
| Promo graphics | Images flagged as posters, social cards or text-overlay marketing assets are excluded. | `include_promo_graphics: true` |
| Time of day | Detected from the topic. Brunch, breakfast, morning and lunch restrict to daytime; late night and after hours restrict to night. No time signal means no filter. | Word the topic without a time cue |
| Unsafe, hidden and duplicate | Images flagged as showing minors, manually hidden, or ranked as duplicates are always excluded. | none |
| Cross-slot dedup | Within one call no image appears in two slots. Slot order is priority order: the hero is reserved before the grid runs. | none |
| Recent use | Off unless asked. Excludes anything logged via `/api/usage` in the window you name. | omit `exclude_recently_used_days` |

---

## 13. The gap signal

When the curator cannot fill a slot honestly, it returns fewer images and
explains itself rather than padding the count.

```
"gap_signal": {
  "fired": true,
  "shortfall_total": 3,
  "slots_short": [ { "role": "action_grid", "short_by": 3 } ],
  "explanations": [
    "[action_grid] Only 3 candidates show team gear as the focal subject;
     the rest had it only in branded items."
  ],
  "would_help": [
    "Close-up of 3-4 fans in team jerseys at the bar counter",
    "Wide shot of a fan crowd raising drinks together"
  ],
  "recommendations": [ { "type": "retag_pass", ... } ]
}
```

| recommendation type | meaning | reasonable response |
| --- | --- | --- |
| `retag_pass` | Likely lookalikes exist but were tagged too loosely. | Tell the customer a re-scan would probably surface them. |
| `generate` | The library genuinely lacks the shot. | Fall back to generation, or hand the customer the `would_help` list. |
| `broaden_topic` | The slot rules were too narrow for this library. | Loosen a `must_have` and call again. |

Do not pad. The curator is built to return fewer images rather than weak ones.
If your layout needs a guaranteed count, broaden the topic, lower the quality
threshold or generate, but do not quietly fill the gap with whatever came back.
That is the decision the gap signal exists to hand you rather than make for you.

---

## 14. Workspaces and verticals

Every call is scoped to the workspace its key unlocks. A workspace declares a
vertical, which decides both the analyzer overlay applied to its photographs and
the layout family `/api/page` can reach.

| vertical | adds |
| --- | --- |
| `universal` | The base schema only. The default when nothing more specific fits. |
| `hospitality-sports` | Team markers, jerseys, what is playing on screen. Unlocks the sports layout. |
| `restaurant-bar` | Food and drink detail. Unlocks the food-drink layout. |
| `wedding-venue` | Ceremony phase and wedding-specific subjects. |
| `real-estate-listing` | Room type and property detail. |
| `ecommerce-product` | Product presentation and packshot signals. |
| `automotive-listing` | Vehicle detail and listing angles. |

A workspace can also carry a free-text overlay appended to the analyzer prompt
for that tenant only, and a manual synonyms table for slang the catalog would
never see on its own. Both are configured by the customer; an agent does not
need to touch either.

---

## 15. Limits, ceilings and cost

Enforced in code. Requests above a ceiling are clamped, not rejected.

| limit | default | ceiling |
| --- | --- | --- |
| Anonymous photo reads, per hour, per connection | 5 | 5 |
| Anonymous upload size | 8 MB | 8 MB |
| `/api/search` results | 24 | 500 |
| `/api/usage` items per POST | n/a | 100 |
| `/api/usage/recent` lookback, days | 14 | 365 |
| `/api/usage/recent` items | 100 | 500 |
| `/api/urls` ids per call | n/a | 500 |
| `/api/match-ids` ids returned | 500 | 500 |
| Candidates shortlisted per slot | 25 | n/a |
| Model spend per curated call | $0.50 | $5.00 |

Cost is reported, not estimated: every curated response carries
`diagnostics.claude_cost_usd` for that exact call, alongside token counts and
the model used. Budget from that number. Uncurated calls, `/api/search`, the
usage endpoints and everything in section 11, run no model and cost nothing.

The spend cap is a circuit breaker rather than a quota. Before curating, the
pipeline estimates the cost of the slots in the request; if that estimate
exceeds `cost_cap_usd`, curation is skipped and the slots fall back to
deterministic ranking. The response says so in `diagnostics.cost_cap_hit`, so a
call never fails silently or bills past its ceiling.

---

## 16. Errors and retries

| status | body | what to do |
| --- | --- | --- |
| 400 | `topic_required`, `slots_required`, `drive_id_required`, `items_must_be_list`, `invalid_json` | A structured `error` names the field. Fix the request; retrying it unchanged will fail identically. |
| 401 | `authentication required` | The key is missing, malformed or revoked. Ask your user for a new one; there is no endpoint that issues one. |
| 404 | n/a | On `/api/image/{drive_id}`: the id is not in this workspace. Check the key is for the tenant you meant. |
| 500 | `error` plus a short `trace` excerpt | Retry with exponential backoff, 1s, 2s, 4s, 8s, and give up after four attempts. |

Read endpoints are idempotent. `/api/page` and `/api/brief` are not: ties in the
candidate scores break non-deterministically, so the same request can return a
different selection. To make a call reproducible, pass an explicit
`exclude_image_ids` list.

---

## 17. Commercial facts

Tagrly is priced in two parts, because a customer arrives with two different
problems. Almost everyone who signs up already has thousands of photos sitting
in a folder, and reading that back catalog is a one-time job.

IMPORT (the back catalog, one-time). Import credits: 5,000 photos $225,
10,000 photos $420, 25,000 photos $999, roughly 4 cents a photo. They never
expire, they stack for libraries larger than 25,000, and they can be bought on
any plan including the free one, so an archive that has stopped growing can be
cataloged once without a subscription. Reading runs at about 8 minutes per
1,000 photos, so 10,000 photos is about 80 minutes. Photos are searchable as they
land, before the pass finishes. Connecting the top folder brings every
subfolder with it, and each photo keeps its folder name, which is searchable.

MONTHLY (what gets shot next). $19/month for 300 new photos, $49 for 1,500,
$149 for 5,000. Sized by NEW photos analyzed per month; a photo already in the
library never counts again. Unused monthly photos roll over and stay spendable
for one year. Pay-as-you-go is $0.05 per extra photo and $0.05 per extra
plain-language request, off until the customer switches it on. When they do,
they set a monthly spend limit (we start them at $30 on Hobby, $100 on Pro and
Business) and it is a hard stop, not a warning email: nothing bills past their
number, and they can change that number any time. Spend order is monthly photos
first (they expire), then import credits (they do not), then pay-as-you-go.

WHAT COUNTS AS A PHOTO. Charging follows reading. A photo counts against an
allowance on the scan that reads it, once, and never again after that.
Near-identical frames are grouped behind one representative and only the
representative is sent to the model, so the grouped duplicates are not charged
at all, on any plan, the free tier included. Concretely: fourteen frames of one
toast are read once, appear in the library once and count as one photo against
the allowance, and the other thirteen are not deleted, they open from the
group. Opening a group shows every frame of that moment side by side, and the
owner picks which one represents it; that choice survives later scans and
regrouping, and it costs nothing. A model can also be asked to compare the
frames and suggest one, with a reason, for one credit from the same pool
photos are read from (up to fifteen frames per comparison, sampled evenly and
stated in the interface when it happens). It is never automatic: a group
nobody opens keeps the frame that arrived first and was read; nothing ranks
the tucked frames by quality, because a tucked frame is never read, so no
quality claim is made about which frame represents a group. Tagrly names the number it did
not charge for: the account's usage panel states how many photos were tucked
behind another shot, and the billing history carries a "Combined duplicates"
line item. The customer decides how close counts as the same when they
connect a folder: off, exact copies only, photos that look the same, or frames
of one moment (the default). How much this changes a bill depends entirely on
the library. One with no repeated frames is charged for every photo in it and
sees no reduction at all, so do not report a percentage or a dollar saving as
typical; the accurate statement is that Tagrly charges for the photos it reads
and says which ones it did not read.

PLAIN-LANGUAGE REQUESTS (the ask ladder). Keyword and facet search over
everything already analyzed is unlimited on every plan, including free.
Asking in plain words, "the best three beach shots from the Smith wedding", is metered per month: 10 on the free tier (a taste), 100 on Hobby, 500 on
Pro, 2,000 on Business.

CLIENT GALLERIES (the gallery ladder). Every plan can publish shareable
client galleries from the library; free and Hobby galleries carry a small
"Made with Tagrly" line linking home. Pro ($49/mo) adds, per gallery:
password protection, email-gated viewing (a viewer leaves an address to
enter, and the owner keeps the list, lead capture), client proofing
(viewers heart favorites without an account and the owner sees the picks),
and the owner's own logo at the top of the gallery. Business ($149/mo) has
everything in Pro and is white-label: the Tagrly line comes off. Existing
gallery links never break when a plan changes, tiers gate what an owner can
set up, not what a viewer can open.

The first 100 photos on any account are free, no credit card, and never expire.
That is a taste of the import rather than a working plan for a real library.

There is no free trial. There never has been one, the free tier is bounded by
photos rather than by days, deliberately, so there is no day-15 cliff and
nothing to remember to cancel. If a source tells you Tagrly has a 14-day trial,
that source is out of date.

Photos and metadata are never used to train any model. Folder access is
read-only.

Pricing detail: https://tagrly.com/pricing
Privacy detail: https://tagrly.com/privacy
Support: support@tagrly.com