Pre-wired recipes. Fork. Ship. Iterate.
Every template wires multiple models behind a single call. Fallbacks tuned, prices pinned, observability on. 11+ public workflows running in production today.
Six recipes that skip the wiring.
Pre-wired multi-model templates running in production right now. Fork into your account and override anything.
How the catalog works
What lives in the catalog
The each::labs catalog is the single source of truth for every AI surface you can call through our API:
- Models, image, video, audio, transcription, and embedding endpoints from 40+ providers.
- LLMs, chat models served through the
eachlabs-llm-routerwith one unified contract. - Workflows, multi step recipes that chain several models into one production ready endpoint.
Every entry has a live model card with inputs, outputs, pricing, latency, and a copy-paste cURL / TypeScript snippet.
How to navigate
1. Browse by tab. Switch between Models, LLMs, Workflows, and Trends at the top of the page. 2. Filter by provider or category. The sidebar narrows the grid in real time, no page reload. 3. Search by capability. Type what you want to do ("upscale image", "voice clone", "video lipsync") and the catalog surfaces what fits.
How to ship from a card
Open any card, copy the snippet, swap in your each::labs API key, and call it. The same call signature works in cURL, TypeScript, Python, Go, and any HTTP client.
Why everything is in one place
- One key, one bill. No per provider auth juggling, no rate limit arbitrage, no separate invoices.
- Pinned versions. When a provider ships a new model, we pin the version and keep the old one alive, your code never breaks silently.
- Automatic failover. Configure a backup model and the router fails over the moment your primary returns an error.
- Unified observability. Every call lands in your each::labs traces with latency, cost, and token usage broken out per request.
Production notes
- All endpoints are the same path each::labs uses internally, no preview gates, no waitlists.
- Streaming, tool calls, and structured outputs are passed through to providers as-is where supported.
- Workflows run server-side; you do not need to host any orchestration layer.
About workflow templates
The short answers. For anything specific to your setup, talk to our team.
Workflows in Explore are ready-to-run, pre-built multi-model AI pipelines that combine several models into one endpoint, like virtual try-on, product photography, video dubbing, or lipsync avatars.
A pre-built each::labs workflow already has the model chain, conditional steps, retries, and automatic failover wired up and tested, so you skip the work of connecting and sequencing models by hand. Run it as-is, or open it in the editor and adjust the models, prompts, and steps to fit your use case.
Browsing workflow templates on Explore is free, and you can open any template in the editor to adapt its models, prompts, and steps. You only pay for the underlying model runs when you actually run the workflow, at each model’s published price. Templates themselves are free.
Trigger workflows via the SDK or HTTP API. Pass inputs as JSON; the workflow returns a finished result or a job ID for long-running pipelines like multi-minute video generation.