Batched Prompt Runner (BYO-key)

What this tool does

When you want to apply one prompt to many inputs — summarise 50 reviews, classify 30 support tickets, generate taglines for a product list — running them one at a time in a chat window is painfully slow. This tool turns a single prompt template plus a list of values into a batch of API calls and runs them in parallel against your own OpenAI or Anthropic key, showing each result in its own row.

Everything happens in your browser. Your key is used only for the direct provider request and is never stored.

How it works

You write a template containing a {{variable}} slot, for example:

Classify the sentiment of this review as positive, negative, or neutral: {{review}}

Each non-empty line in the values box is substituted into that slot to produce one prompt. The runner then dispatches requests with a small pool of workers — you choose between one and five in flight at once — so the batch finishes quickly without tripping provider rate limits. Each row updates independently to running, done, or error, and failed rows show the provider’s error message so you can see exactly which input broke and why.

Practical worked example

Suppose you have a CSV with 40 product names and you want a one-sentence marketing tagline for each. Export the product names to a plain text file, one per line. Your template might be:

Write a single punchy marketing tagline for this product: {{product}}
Keep it under 12 words and focus on the key benefit.

Paste the 40 product names into the values box, set concurrency to 3, choose your model and key, and click Run. The table fills in with 40 taglines, each tied to its input row. The whole run finishes in a minute or two rather than the 20-plus minutes it would take copying and pasting one at a time into a chat window.

Tips and limits

• Keep concurrency at 2 for free-tier or low-rate-limit keys, and raise it only if you are not seeing 429 errors.
• Put any fixed instructions (format, tone, length) in the template itself so every request is consistent — the only thing that should change between rows is the substituted value.
• For genuinely large jobs (thousands of rows), use the provider’s server-side Batch API instead: it is roughly half the cost and designed for volume. This tool is the fast interactive option for everyday batch work.

Common use cases

Classification at scale. Feed a list of support tickets, reviews, or customer messages to a classification prompt and get a labelled table back in minutes — without writing any code or managing API credentials on a server.

Data extraction. If you have a list of raw text snippets (addresses, product descriptions, email signatures) and want a specific field extracted from each, a template like Extract the city name from this text: {{text}} runs across every row.

First-draft generation. Blog intros, meta descriptions, email subject lines, or social posts for a product list — prompt once, run everywhere in the list, then edit the outputs you want to keep.

Consistency checks. Run the same question against multiple passages of text to see if your model answers consistently, which is useful during prompt development and evaluation.