Ellf: developer tools for developers

What it does

• Coding assistant integration with Claude Code, Codex, GitHub Copilot and others
• Annotation platform built on Prodigy, plus LLM-powered auto-annotation agents
• A data-private cluster on your machine or in your cloud; Explosion says your data is never sent to it
• Project plans that log strategy, decisions and results
• Project templates such as support-ticket extraction and LLM pipeline evaluation

• Cursor: an AI coding workspace
• Claude: a general LLM via API
• ChatGPT: a general LLM via API

Editorial note

Ellf, from Explosion, the makers of spaCy and Prodigy, is a virtual NLP engineer that plugs into coding assistants such as Claude Code, Codex or GitHub Copilot. It helps them plan, annotate, train and evaluate custom NLP models on infrastructure you control. A free Solo plan covers one person on one machine.

Who it’s for

• Developers replacing a slow or costly LLM API call with a task-specific model
• Teams with private or regulated data that must stay on their own infrastructure
• Data teams that need quick annotation and evaluation sets
• Not ideal for: non-technical users; you work through a coding assistant and a CLI.

Key use cases

• Build apps & code

What you can do with it

Explosion says coding assistants fall short on data-centric NLP work in three ways: they cannot get human feedback on real data, they are impractical for long-running jobs, and they default to “just use an LLM.” Ellf adds annotation with Prodigy and LLM auto-annotation agents, execution on a Kubernetes cluster you run, and a knowledge base from more than 10 years of applied NLP.