Route to the best free model.
Fall back only when needed.
OptiSift plugs directly into Hermes, OpenClaw, and any OpenRouter-compatible agent. It tries every free model first, enforces a hard monthly spend cap, and only touches paid tiers when free ones fail — one env var to wire it up.
Fast, code-specialized free models first
Routes to Groq Llama and DeepSeek Coder free tiers — both optimized for instruction-following and code generation at zero cost.
Routing through providers
The complete routing stack
Everything you need to run free-first LLM routing in production.
Without babysitting model availability, managing spend, or writing fallback logic yourself.
Policy chain executor
CoreOrdered tier waterfall: free → floor → paid. Advances on 429, 403, 5xx, timeout, or empty stream. Up to 3 failover attempts per request.
Streaming SSE passthrough
CoreFull text/event-stream proxy with mid-stream failure detection. If a free model drops the stream early, OptiSift retries the next tier without the client disconnecting.
Hard spend caps
CostPer-key and per-org monthly USD limits checked before any paid tier. Never silently overspend. Free models are always uncapped regardless of the limit.
Model health tracking
Consecutive failures put a model in a cooldown window. The routing engine skips unhealthy providers without DB round-trips — in-memory, process-local, auto-recovering.
Request logs & spend ledger
Every request logged: model used, tier, fallback chain, tokens, estimated cost, latency. Query /admin/spend or check the dashboard.
BYOK support
Attach your own Anthropic, OpenAI, or Groq keys per policy. Paid calls route through your provider accounts — no OpenRouter markup on direct calls.
OpenAI-compatible drop-in
Drop-inPOST /v1/chat/completions speaks the exact same schema. Change OPENAI_BASE_URL and the default model slug — that's the entire client change required.
5 routing modes
cheapest · balanced · quality · fastest · free_only. Set per-request via the virtual model slug or baked into the policy chain config.
Model change alerts
OptiSift probes 400+ free models every 15 minutes. Get notified by email when a free model goes paid, dies, or a new one appears.
Pricing
Free to watch. $9/mo to never worry.
Watch
- Live status for 377+ models
- Heatmap & funnel views
- 3 alert subscriptions
- 7-day status history
- Live status dashboard
- 3 model alert subscriptions
- Weekly status digest email
- Public vertical rankings
Pro
- Everything in Watch
- Unlimited alert subscriptions (email, webhook, Slack)
- Failover proxy with auto-routing
- 100 failover saves per month
- 30-day ranking history
- 7-day free trial
- Unlimited alert subscriptions
- Failover proxy with auto-switching
- 100 model saves/mo
- Full status history & export
- Priority support
Team
- Everything in Pro
- Up to 10 API keys
- 500 failover saves per month
- All vertical rankings + full history
- Leaderboard API access
- CSV export & priority support
Founding Member
- Lifetime Team features — no monthly fee
- 1,000 failover saves on signup
- Name in product credits
- Locked-in pricing, forever
- Influence the roadmap
Frequently asked questions
Everything you need to know about OptiSift and free-tier LLM tracking.
Stay in the loop
Weekly digest of free-model status changes, new additions, and ranking shifts.