Can an AI assistant create product photos for my whole catalog?

Not by itself. Claude does not generate photos, and an assistant that can make images, such as Codex, makes them inside the conversation from the photos and instructions you give it there. That works for a handful of images. It does not give a catalog a saved record of each product, a reusable shoot setup, a price before you commit, or a job that keeps running when you close the chat.

Connect Nightjar to Claude, Claude Code or Codex and the assistant can hand catalog work to a production system instead. It plans a Nightjar batch across up to 100 of your saved Products using one Recipe's settings, shows the total Credit cost and a spending ceiling before anything runs, and asks you to approve a few sample images. Nightjar then produces the rest, even after you close the chat, and every image lands in your Team Library under its Product.

What happens if you ask an assistant on its own?

None of these assistants is built to photograph a catalog alone, and each stops at a different point.

• Claude on the web, desktop, Cowork or mobile: Anthropic's help center says "Claude doesn't generate photos or illustrations the way image-generation tools do". Without a connected image tool, asking it to photograph your catalog gets you a plan, listing copy or diagrams, not product photos.
• Claude Code has the same limit, but it can write and run code that calls a separate image service. That can work, and it makes you the owner of a pipeline: keys, uploads and downloads, retries, cost tracking, and keeping each product's reference photos and settings consistent from one run to the next.
• Codex can generate and edit images itself, including from reference images, inside the conversation. OpenAI's Codex image generation guide says image turns use included limits 3 to 5 times faster on average than similar turns without images, and points larger batches to API-key billing. The guide describes making and refining images in the chat. It does not describe a saved product record, a reusable shoot setup, or a batch with a sample checkpoint and a spending ceiling.

Where images do come out of a chat or a script, they still have to be collected, renamed and moved to wherever the rest of your catalog work lives.

What changes when Nightjar is connected?

Nightjar connects through MCP, the open standard AI assistants use to connect to other tools. You add the Nightjar connector once, sign in to Nightjar in the browser and choose your Team. There is no API key, and connecting is free. From then on the assistant can read your Team's saved Products and Recipes, start Nightjar work and show you the results. The assistant plans and talks; Nightjar makes the images, with the same Products, Recipes, Credits and Library you use in the web app. Images cost your Team's normal Credits at the same prices, and the assistant states the cost before it submits paid work.

The connector works in Claude on the web, the desktop app, Cowork and mobile as a custom connector, and in Claude Code and Codex. On Claude Team and Enterprise plans, an Owner adds Nightjar in Organization settings first. Making product photos in Claude with Nightjar shows what the conversation looks like, and the Nightjar MCP setup guide has the steps for each client, plan requirements and current limits.

How do you photograph a catalog from your assistant?

1. Save the Products first. A catalog batch runs across saved Products. A Product is Nightjar's record for one visually distinct item: several Product Photos plus a factual description and optional dimensions, so every image starts from more evidence than one loose upload. Your assistant can build them with you. It can import a Product from a public product page, showing you the details and photos to review before anything is saved, or save photos you upload. Imports are reviewed one page at a time.
2. Ask for the batch in plain language. For example: "Apply our Summer Recipe to every product in the Linen collection. Show me three samples first." A Recipe is your Team's saved shoot setup: Photography Style, background, Framing or Pose, Fashion Model, Custom Directions and output settings. The assistant reads it and applies its settings to every item. Nightjar doesn't know your store's collections, so name the Products or let the assistant match them from your saved list. One batch holds up to 100 items; a larger catalog runs as several reviewed batches.
3. Review the free plan. Planning costs nothing and starts nothing. The plan lists the Products, the settings, the images each Product gets, and the Credit cost of the sample and of the remainder. It also carries a Credit ceiling: Nightjar rejects a plan that would cost more, and the ceiling is a maximum, not a target. Each Product gets its own separate Generation, so you receive images of each item rather than one group shot.
4. Approve the samples and check them. Nightjar runs a few representative Products first and holds the rest until they finish. Compare the samples with the real products. This is the point to adjust the setup, before most of the Credits are spent. Approving a phase reserves only that phase's cost.
5. Approve the rest, then close the chat if you like. The batch runs in Nightjar, not in the conversation. Results land in the Team Library, attached to each Product, where your teammates can see them too.
6. Pick it up from any new conversation. Ask your assistant for your recent batches to check progress. You can pause a batch, which holds unstarted items and releases the Credits held for them; resume it, which reserves Credits only for unstarted items and never reruns finished ones; or stop it, which skips what hasn't started and keeps every finished image. Work already running finishes in each case. If you change a Product's photos or facts mid-batch, Nightjar pauses the batch's unstarted work until you review a new plan for what's left. To end a batch, ask your assistant to stop it; removing the connector does not stop a batch.

For a sense of scale, take a hypothetical collection of 60 Products with one 2K image each. At the current price of one Credit per completed 1K or 2K image, the plan totals 60 Credits. With three samples, approving the sample reserves 3 Credits and approving the rest reserves the other 57. Your own total depends on your settings and the prices at the time.

Why use Nightjar for catalog photos in an assistant?

With Nightjar connected, your assistant plans and Nightjar produces. The assistant turns "photograph the Linen collection with our Summer look" into a Nightjar batch: a separate Generation for each saved Product with the same Recipe settings, a free plan with the total Credit cost and a ceiling Nightjar enforces, and a few samples to approve before the rest runs. You keep two decisions, the price before anything runs and the sample before most of the money is spent. You don't re-upload photos, re-explain the brand for each product, keep a chat open while work runs, or move files between tools.

Several product-photo tools now offer assistant access, so the connection alone isn't the reason to choose Nightjar. The reason is what sits behind it. Products store the evidence for each item. A Recipe stores how you photograph them, so the next collection and every teammate start from the same setup. The batch adds a cost plan, a sample checkpoint, a ceiling and recovery from a new conversation. The images come from the same Nightjar system as the web app, including built-in visual review that can retry obvious failures at no extra Credit cost.

When should you use the API instead?

The connector is for a person steering the work through an assistant. If your catalog system should produce images on its own, on a schedule or whenever a product is added, with nobody in a chat, use the Nightjar API. Your software owns the loop and calls Nightjar directly, and API Access comes with any paid plan. Starting Product Photography through the Nightjar API covers the first request, and generating AI product photos in bulk for many SKUs compares the web app and API methods.