classifier-dev Classify API

openapi: 3.2.0
info:
  title: classifier.dev Classify API
  version: 1.0.0
  summary: Zero-shot text classification with calibrated confidence. Free without a key; paid plans raise rate limits 10-100x.
  description: 'Send text and a list of labels, receive the label that fits, a calibrated confidence and a score per label. Up to 1,000 texts per request, ~1s. Tiers: fast (default) and smart, which re-asks answers below 0.7 confidence of a fast reasoning model. Free limits are per IP and counted in classifications: 3,000/min and 20,000/day on fast, 200/min and 2,000/day on smart. Benchmarks: https://classifier.dev/benchmark


    Authentication: optional for classification. Pro ($20/month) workspaces provide fast 30,000/min and 200,000/day, smart 2,000/min and 20,000/day shared across workspace keys and agents. Partner keys remain supported (see https://classifier.dev/auth.md).


    Versioning: the current major is v1, addressed as POST /v1/classify; POST / is an alias that tracks the current major. Response shapes are additive within a major (fields are added, never renamed or removed). Every response carries an x-api-version header. A breaking change ships as /v2 beside /v1, and /v1 then carries Deprecation and Sunset headers (RFC 9745 / RFC 8594) for at least six months before removal.


    Rate limits: RateLimit-Limit and RateLimit-Policy (IETF draft-ietf-httpapi-ratelimit-headers) on every classification response, RateLimit-Remaining once the limiter has been consulted (every 200 and 429), Retry-After on 429s. Idempotency: synchronous paid requests reject replayed keys. Whole-document jobs accept an optional UUID Idempotency-Key and return the same job on retry without charging again.


    Errors: classification failures return {error, code}; workspace authorization and billing errors return {error}. See components.schemas.Error for classification codes. The two GET forms answer plain text (`error:`, `usage:`, `try:` lines) unless ?verbose=1 or Accept: application/json asks for the JSON object.


    MCP: the same capability as tools at https://classifier.dev/mcp (Streamable HTTP, no auth), documented at https://classifier.dev/mcp-setup. Batch: the inputs array is the batch operation — up to 1,000 texts per request; POST /v1/classify/batch is an alias for callers that look for one.'
  contact:
    name: Michael Ryaboy
    url: https://cal.com/michaelsf/coffee
    email: contact@classifier.dev
  license:
    name: MIT
    url: https://github.com/mrmps/classifier-dev/blob/main/LICENSE
  termsOfService: https://classifier.dev/terms
  x-api-versioning:
    scheme: url-path
    current: v1
    aliases:
    - /
    - /v1/classify
    - /v1/classify/batch
    deprecationSignal: Deprecation and Sunset response headers, at least 180 days notice
    changelog: https://github.com/mrmps/classifier-dev/commits/main
  x-mcp:
    url: https://classifier.dev/mcp
    docs: https://classifier.dev/mcp/docs
    card: https://classifier.dev/.well-known/mcp/server-card.json
servers:
- url: https://classifier.dev
security:
- {}
- partnerKey: []
tags:


# --- truncated at 32 KB (95 KB total) ---
# Full source: https://artifacts.apis.io/classifier-dev/openapi/classifier-dev-classify-api-openapi.yml