classifier.dev has launched an official CLI tool to bring fast, keyless zero-shot text classification directly into Unix shell pipelines.
Zero-shot classification service classifier.dev has released an open-source command-line interface (`classifier-dev`), allowing developers to classify and filter text directly from the terminal without API keys or accounts. The underlying platform provides fast zero-shot labeling over plain HTTP, designed to pre-filter search results, streams, and log feeds before they consume expensive LLM context windows. With the new `classify` command, users can stream plain text, JSON, or NDJSON inputs via shell pipes, process batches of up to a thousand items per request, review uncertain predictions based on confidence thresholds, and generate label count histograms.
Bringing keyless zero-shot classification to standard Unix pipes makes LLM context pre-filtering and data triage frictionless for developer and agent pipelines.
- –Zero-friction developer ergonomics: By requiring no account or authentication, the CLI can be installed and immediately embedded in local scripts, pipelines, and CI jobs without credential management.
- –Context window savings: Acting as an upstream filter allows systems to drop irrelevant text cheaply before passing high-signal candidates to expensive reasoning models.
- –High-throughput stream processing: Native streaming and batching mean piping massive files directly to shell tools like `head` or custom consumers works out of the box with automatic retries.
- –Multi-surface integration: Adding a CLI alongside existing Model Context Protocol (MCP) servers and agent skill discovery solidifies classifier.dev as a lightweight, cross-environment routing utility.
DISCOVERED
1h ago
2026-09-19
PUBLISHED
2h ago
2026-09-19
RELEVANCE
AUTHOR
michael_chomsky