Roboflow

Roboflow is the most accessible computer vision development platform — covering the full CV pipeline from dataset management to training to inference deployment. For agents: upload images for a CV dataset, trigger training jobs, run inference against hosted models, and integrate vision capabilities into pipelines. Large public model/dataset library (Universe) accelerates development. Strong free tier for individual/small-team use. Confidence is docs-derived.

REST API: Inference — POST https://detect.roboflow.com/{model}/{version}?api_key={key} with image (multipart or base64). Returns predictions: bounding boxes, class labels, confidence scores. Dataset management API: upload images, create/manage datasets and versions. Python SDK: from roboflow import Roboflow; rf = Roboflow(api_key=...); project = rf.workspace().project("my-model"); model = project.version(1).model; pred = model.predict("image.jpg", confidence=40).json(). Inference available as hosted API or self-hosted inference server (Docker).

API key auth: api_key query parameter or RF SDK initialization. Keys from Roboflow Dashboard. HTTPS enforced. No OAuth2. Public models: accessible with any valid key. Private models: key must be from the owning workspace. Rate limiting: per-key and per-model limits (plan-based).

Inference returns 200 with predictions even on empty detections ({predictions: []}). 401 for invalid key, 403 for private model access, 404 for model not found, 429 for rate limit. Training jobs are async (status polling or webhook callback). Roboflow Cloud uptime at status.roboflow.com. Self-hosted inference server (Roboflow Inference) available for air-gapped or high-throughput deployments.

docs.roboflow.com covers quickstart, dataset upload, annotation, training, inference API, and deployment guides. Python/JavaScript SDKs. Getting started: sign up (free), create a project, upload 10 images, train a model in under 20 minutes (free GPU time available). Public dataset library (Roboflow Universe) has millions of labeled images. Community via Roboflow Forum and Discord.