Integrations · Cursor
Cursor with the lowest-cost frontier coder we've found.
Cursor talks OpenAI. Routify talks OpenAI. Override one base URL and you're sending every Tab, Composer, and Agent request through DeepSeek V3.2 — at $0.14 per million tokens, roughly 18× cheaper than GPT-4o for code. No proxy, no plugin, no patches.
Setup in 30 seconds
Cursor uses an OpenAI-compatible base URL. Override OPENAI_BASE_URL and you’re done — no plugin, no extension, no proxy.
💡 Pin a fallback chain in Routify dashboard so when DeepSeek throttles, your Cursor request transparently routes to Kimi or Qwen3 Coder.
# In Cursor → Settings → Models, add a custom OpenAI provider:
# API Base: https://routify.bytedance.city/v1
# API Key: rtf_xxx_your_routify_key
#
# Or set env vars and restart Cursor:
export OPENAI_BASE_URL=https://routify.bytedance.city/v1
export OPENAI_API_KEY=rtf_xxx_your_routify_key
# Then in Cursor, type a model id like:
# deepseek-chat (cheapest, GPT-4-class)
# claude-opus-4-7 (frontier, pass-through pricing)
# kimi-k2 (long-context tool use)Cursor caches the model list — you may need to restart Cursor after switching base URL.
Profiles
Picked for Cursor.
Three battle-tested model combos. Pick the profile that matches your priority and copy the model id straight in.
Cheap
You write a lot of code, you don't want to think about tokens. DeepSeek V3.2 ships strong code at 1/100 of GPT-4o pricing. The fallback chain catches the 1% of cases where DeepSeek is overloaded.
Cost note
Median Cursor session ≈ $0.04. A heavy day under $1.
Primary model
Fallback chain
Smart router auto-fails over to the next id if the primary is over budget or down.
FAQ
Will Cursor Tab still work?
Yes. Cursor Tab uses a smaller proprietary model that doesn't go through OPENAI_BASE_URL. Composer / Agent / chat-with-codebase all flow through Routify.
Does Cursor know it's talking to a different model?
No — to Cursor it looks like any OpenAI-compatible model. We translate Anthropic-only features (cache_control, anthropic-beta headers) on the way through.
How fast is DeepSeek V3.2 on Routify?
Direct upstream is ~30-60 t/s. Routify adds < 80ms p95 routing overhead on top. We auto-fail over to a backup channel if the primary slows below the SLO.
Can I bring my own Anthropic key?
Yes — coming Q3 2026. You'll attach your own key in dashboard and pay only the routing fee (0% on first 1M requests/month).
Does this work on Mac / Windows / Linux?
All three. The base URL setting lives in Cursor itself — Routify is platform-agnostic.
Ship with Cursor in 30 seconds.
$5 free credit. No credit card required.