Auto-routing AI API
ElseLane
If the primary fails, take the else lane. One key, one prompt — we classify, route, and fail over across models so your app keeps answering.
curl https://elselane.com/api/v1/answer \
-H "Authorization: Bearer el_live_..." \
-H "Content-Type: application/json" \
-d '{"prompt":"Refactor this function for clarity"}'Built for reliability, not a model zoo
Provider outages, rate limits, and model pickers slow indie and SMB apps down. ElseLane is a thin reliability layer: send the prompt, get an answer that gets through.
Without ElseLane
Hard-code a provider. When it fails, your users see errors. Or you build failover yourself and maintain keys, chains, and billing across vendors.
With ElseLane
One key. Task-based chains. Circuit breakers. Prepaid credits. Transparent attempts so you can see exactly which lane answered.
How it works
Three steps between your prompt and a durable answer.
01
Classify
We inspect the prompt and pick a task type — chat, code, reasoning, and more — without you choosing a model.
02
Route
Each task type has a provider chain tuned for cost and quality. Your request hits the primary first.
03
Fail over
If the primary errors, times out, or is circuit-open, we take the else lane automatically and keep answering.
What you get
Simple reliability for apps that need answers, not a 400-model marketplace.
Prompt-only API
POST a prompt. Optional system message. No model picker required for day-to-day use.
OpenAI-compatible shim
Drop in /api/v1/chat/completions with model: "auto" when you already speak OpenAI.
Prepaid credits
Buy $10 / $25 / $50 packs. Usage billed at estimated provider cost × 1.10 — no surprise invoices.
PII guardrails
High-risk secrets and identifiers are blocked (or redacted) before providers see them.
Transparent attempts
Responses include which provider and model answered, whether failover ran, and each attempt.
Metadata-only storage
Prompts and answers are not retained. We keep usage metadata for billing and debugging.
Simple credit packs
1 credit = $0.0001. Successful answers are charged at estimated provider token cost × 1.10 (10% platform fee), rounded up. No monthly subscription required.
Verified accounts receive a small starter credit balance. See docs for billing details.
Trust by default
Prompts not stored
Request and response bodies are not retained after the call completes.
Usage metadata only
Status, model, provider, credits, and latency for dashboards and support.
Export or delete
Export usage CSV or delete your account from the dashboard anytime.
FAQ
Quick answers. Full list on the FAQ page.
Do I pick the model?
No. Send a prompt; ElseLane classifies and routes. Use model: "auto" on the OpenAI-compatible endpoint.
What if a provider is down?
We fail over along the task chain. You get an answer when any lane succeeds, plus an attempts list.
Are prompts stored?
No. We store usage metadata only (status, model, credits, latency).
Keep the else lane open
Create an account, grab an API key, and ship answers that survive provider hiccups.