Is my data really private?

Yes. All substitution runs in your browser tab. No data is sent to any server. Session history lives only in localStorage. Shareable URLs encode your data as a base64 string in the URL — no server stores it. Verify by opening your browser's Network tab (F12 → Network) and observing zero outbound requests when you click "Process All".

How does batch pagination work?

Set a batch size (10, 25, 50, 100, 250, or 500) using the Batch Size control above the Process button. "Process All" runs the first batch starting from row 1. A "Process Next N" button then appears — click it to advance through the dataset in controlled chunks. This is useful for spot-checking results before committing to a full large-dataset run.

How do I save a custom template?

Click the "Save" button next to the template editor header. Give your template a name and optional description, then click "Save Template". Your custom templates appear in the Template Library under the "My Templates" tab. They are stored in your browser's localStorage and persist across sessions. You can edit or delete them at any time.

Diff view shows a side-by-side comparison of the original template (left) and the substituted result (right). {{variable}} placeholders are highlighted in amber on both sides — the left shows the placeholder name, the right shows the substituted value. This makes it easy to audit substitution accuracy at a glance, especially for multi-variable CSV templates.

How does multi-column CSV substitution work?

Upload or paste a CSV where the first row contains column headers (e.g., product_name, category, price). The tool automatically maps each {{column_name}} placeholder in your template to the corresponding column value for each row. Column names are case-insensitive and spaces are converted to underscores.

How accurate is the token estimator?

The estimator uses the ~4 characters per token heuristic, accurate to within 10–15% for standard English text. For non-English content, code, or heavily punctuated text, actual token counts may differ. For precise counts, use the tokenizer provided by your target model's API.

About Us

We built the tool we wished existed

PromptProcessor started as an internal tool. We were running the same prompt against hundreds of product names, and the existing options all had the same problem: they required an account, sent data to a server, or charged per row. We wanted something faster, simpler, and private.

The story behind it

The first version was a 50-line JavaScript snippet that did one thing: replace a {{variable}} in a text box with each line from a textarea. It was ugly, but it worked. We shared it with a few people and the response was immediate — everyone had the same problem.

Over the following months, we added multi-column CSV support, a template library, session history, the diff view, the token estimator, and eventually the real AI runner. Each feature came from a real workflow need — either our own or from people who reached out after using the tool.

The design philosophy stayed constant throughout: everything runs in the browser, nothing gets stored on a server, and the core tool stays free. That is not a marketing position — it is a constraint we set at the start and have never compromised on.

Who uses it

The tool gets used by a wider range of people than we expected. E-commerce teams use it for product descriptions. Marketing teams use it for email personalization. SEO agencies use it to generate meta descriptions at scale. Developers use it for commit messages and documentation. ML engineers use it to build annotation instructions.

What they all have in common is a list of inputs and a consistent task to perform on each one. The specific domain does not matter much — the pattern is the same, and the tool handles it the same way regardless.

We also hear from prompt engineering educators who use it in courses, and from individual researchers who process datasets they cannot share with third-party services. The privacy guarantee matters differently to different people, but it matters to all of them.

What we believe

These are not aspirational values — they are the decisions we made when building the tool and have continued to make as it has grown.

Privacy is non-negotiable

We built the tool so that it literally cannot collect your data. There is no server-side processing. There is no database storing your prompts. The architecture enforces the promise.

Speed matters

Running in the browser is not a compromise — it is a feature. No API round-trips means no latency. A 500-row batch processes in under a second. That changes how you think about the tool.

Free means free

The core tool is free with no account, no credit card, and no expiry. We believe the most useful AI productivity tools should be accessible to everyone, not just teams with budget.

Transparency in how it works

The token estimator, quality scorer, and diff view are not black boxes. We document how they work so you can understand and trust the output — not just accept it.