One API for AI.
Models · Search · Documents · Speech
Build AI products without stitching together AI vendors. One API key, one bill, normalized responses, recommended routing and automatic failover across the infrastructure your application needs.
from monokey import Client
ai = Client(api_key="mk_live_...")
ai.generate(messages=[{"role": "user", "content": "Explain duration risk."}], optimize="auto")
ai.search(query="latest NVIDIA AI chip announcement")
ai.parse(file="10k.pdf", output="markdown")
ai.transcribe(file="meeting.mp3", diarization=True)Models
OpenAI-compatible /v1/chat/completions plus a native generate() with streaming, tools and JSON schema.
Search
Navigational, factual and research queries routed to the right web search upstream.
Documents
PDF and image parsing. Digital PDFs use free local extraction; scans go to OCR.
Transcription
Prerecorded audio with diarization and language detection, async for long files.
Recommended routing, not lock-in
Every request is classified, filtered by hard capability requirements, checked against provider health and your constraints, then ranked by quality, cost, latency and reliability. Choose optimize="auto", cheap, fast, quality, or pin a specific model. Set max_cost to cap spend per request.