Entry Point AI

Entry Point AI is a browser-based workspace for turning a prompt that mostly works into a small, task-specific language model. It is built around three tools that share one dataset: Transforms, Synthetic Data and Fine-tuning. You can start with any of them, and together they take a table of raw records through to a trained model you call from your own code.

Transforms, synthetic data and fine-tuning

A transform is a prompt run against every row of an imported table, with columns referenced as template variables, so it fills in a new AI-generated column for tagging, classification, extraction or enrichment. Before you trust it at scale, you can score the output against known answers by exact match or have a separate scoring prompt grade it for subjective work. When labelled examples are scarce, the synthesis tool generates new rows seeded from your real ones, stays within the labels you defined, marks every row as synthetic or real, and lets you approve examples one by one or accept them automatically.

Fine-tuning works through a no-code form: choose a template, a provider and a base model, see the token count and an estimated cost and training time, and optionally open the advanced hyperparameters. Finished models can be compared against each other and against plain prompts, tried in a playground, or shared as a simple hosted test page that logs every completion.

Providers and data

Entry Point sits on top of model providers rather than hosting models itself. You connect your own accounts, such as OpenAI, Anthropic, Google AI, Groq and Together AI for open models, and training and generation are billed by those providers directly. Datasets can be exported as JSONL or CSV at any time, including after cancelling.

Pricing

You can try it free with up to 300 rows. Paid plans include every feature and differ by row limit and seats: Starter is $49 a month (5,000 rows, 3 seats), Growth $99 (25,000 rows, 5 seats) and Pro $249 (100,000 rows, 10 seats, premium support). Enterprise is quoted on request.