Skip to navigation
Vision APIsVision

Vision API

For chat over text, images and short videos, use the OpenAI-compatible Chat Completions API.

The Reka Vision API lets you upload and organize videos, process them on demand, then search them and ask questions about them with AI.

The Vision API pre-processes and stores your videos and embeddings for you. To handle that yourself, use the Chat Completions API with a video input.

Key features

  • Video Management: Upload, retrieve, list, update, and delete videos
  • Video Groups: Organize videos into named collections
  • Video Features: Process videos on demand for transcripts, captions, embeddings, and object detection
  • Video Search: Find timestamped moments with semantic or text search
  • Video Chat: Ask questions about up to five videos at once
  • Video Insights: Read transcripts, captions, scenes, and objects, and detect prompted objects in a segment
  • YouTube Automations: Generate clips whenever a YouTube channel publishes a new video
  • MCP Server: Connect agent clients to Reka Vision for video research workflows
  • Metadata Tagging: Generate tags for videos
  • Highlight Clip Generation: Generate shorter highlight clips from your longer videos

Getting started

  1. Upload a video with POST /v2/videos
  2. Process it with the feature endpoints under /v2/videos/{video_id}/features, using POST /v2/videos/{video_id}/features/plan to see what to run next
  3. Search your videos with POST /v2/search
  4. Ask questions about your videos with POST /v2/chat
  5. Generate tags for your videos with POST /v1/qa/indexedtag
  6. Generate highlight clips from your videos with POST /v1/clips

Vision MCP Server

Connect Claude Desktop, Cursor, Claude Code, and other MCP clients to Reka Vision through the Reka MCP server. Your agent can upload, manage, search, understand, and analyze videos using Reka Vision without calling the API directly.

Authentication

All Vision API requests require authentication using an API key in the X-Api-Key header:

X-Api-Key: YOUR_API_KEY

Base URL

https://vision-agent.api.reka.ai

Healthcheck

Ping /health to check that the service is up.

curl -s https://vision-agent.api.reka.ai/health | jq .
import requests
def health_check():
    url = "https://vision-agent.api.reka.ai/health"
    resp = requests.get(url, timeout=5)
    resp.raise_for_status()
    print("[HEALTH ✓]", resp.status_code)
health_check()

Finetuning

Our engineering team offers finetuning.

Pricing starts at $0.5 per input video minute, output text included, with discounts for larger training runs.